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>
76 #include <sys/ktrace.h>
79 #include <security/audit/audit.h>
80 #include <security/mac/mac_framework.h>
83 #include <vm/vm_extern.h>
84 #include <vm/vm_param.h>
86 #include <vm/vm_map.h>
87 #include <vm/vm_page.h>
89 #include <vm/vm_domain.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, , , 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 reaper_abandon_children(struct proc *p, bool exiting)
131 struct proc *p1, *p2, *ptmp;
133 sx_assert(&proctree_lock, SX_LOCKED);
134 KASSERT(p != initproc, ("reaper_abandon_children for initproc"));
135 if ((p->p_treeflag & P_TREE_REAPER) == 0)
138 LIST_FOREACH_SAFE(p2, &p->p_reaplist, p_reapsibling, ptmp) {
139 LIST_REMOVE(p2, p_reapsibling);
141 p2->p_reapsubtree = p->p_reapsubtree;
142 LIST_INSERT_HEAD(&p1->p_reaplist, p2, p_reapsibling);
143 if (exiting && p2->p_pptr == p) {
145 proc_reparent(p2, p1);
149 KASSERT(LIST_EMPTY(&p->p_reaplist), ("p_reaplist not empty"));
150 p->p_treeflag &= ~P_TREE_REAPER;
154 clear_orphan(struct proc *p)
158 sx_assert(&proctree_lock, SA_XLOCKED);
159 if ((p->p_treeflag & P_TREE_ORPHANED) == 0)
161 if ((p->p_treeflag & P_TREE_FIRST_ORPHAN) != 0) {
162 p1 = LIST_NEXT(p, p_orphan);
164 p1->p_treeflag |= P_TREE_FIRST_ORPHAN;
165 p->p_treeflag &= ~P_TREE_FIRST_ORPHAN;
167 LIST_REMOVE(p, p_orphan);
168 p->p_treeflag &= ~P_TREE_ORPHANED;
172 * exit -- death of process.
175 sys_sys_exit(struct thread *td, struct sys_exit_args *uap)
178 exit1(td, uap->rval, 0);
183 * Exit: deallocate address space and other resources, change proc state to
184 * zombie, and unlink proc from allproc and parent's lists. Save exit status
185 * and rusage for wait(). Check for child processes and orphan them.
188 exit1(struct thread *td, int rval, int signo)
190 struct proc *p, *nq, *q, *t;
193 mtx_assert(&Giant, MA_NOTOWNED);
194 KASSERT(rval == 0 || signo == 0, ("exit1 rv %d sig %d", rval, signo));
198 * XXX in case we're rebooting we just let init die in order to
199 * work around an unsolved stack overflow seen very late during
200 * shutdown on sparc64 when the gmirror worker process exists.
202 if (p == initproc && rebooting == 0) {
203 printf("init died (signal %d, exit %d)\n", signo, rval);
204 panic("Going nowhere without my init!");
208 * Deref SU mp, since the thread does not return to userspace.
210 if (softdep_ast_cleanup != NULL)
211 softdep_ast_cleanup();
214 * MUST abort all other threads before proceeding past here.
218 * First check if some other thread or external request got
219 * here before us. If so, act appropriately: exit or suspend.
220 * We must ensure that stop requests are handled before we set
223 thread_suspend_check(0);
224 while (p->p_flag & P_HADTHREADS) {
226 * Kill off the other threads. This requires
227 * some co-operation from other parts of the kernel
228 * so it may not be instantaneous. With this state set
229 * any thread entering the kernel from userspace will
230 * thread_exit() in trap(). Any thread attempting to
231 * sleep will return immediately with EINTR or EWOULDBLOCK
232 * which will hopefully force them to back out to userland
233 * freeing resources as they go. Any thread attempting
234 * to return to userland will thread_exit() from userret().
235 * thread_exit() will unsuspend us when the last of the
236 * other threads exits.
237 * If there is already a thread singler after resumption,
238 * calling thread_single will fail; in that case, we just
239 * re-check all suspension request, the thread should
240 * either be suspended there or exit.
242 if (!thread_single(p, SINGLE_EXIT))
244 * All other activity in this process is now
245 * stopped. Threading support has been turned
250 * Recheck for new stop or suspend requests which
251 * might appear while process lock was dropped in
254 thread_suspend_check(0);
256 KASSERT(p->p_numthreads == 1,
257 ("exit1: proc %p exiting with %d threads", p, p->p_numthreads));
258 racct_sub(p, RACCT_NTHR, 1);
260 /* Let event handler change exit status */
265 * Wakeup anyone in procfs' PIOCWAIT. They should have a hold
266 * on our vmspace, so we should block below until they have
267 * released their reference to us. Note that if they have
268 * requested S_EXIT stops we will block here until they ack
271 _STOPEVENT(p, S_EXIT, 0);
274 * Ignore any pending request to stop due to a stop signal.
275 * Once P_WEXIT is set, future requests will be ignored as
278 p->p_flag &= ~P_STOPPED_SIG;
279 KASSERT(!P_SHOULDSTOP(p), ("exiting process is stopped"));
282 * Note that we are exiting and do another wakeup of anyone in
283 * PIOCWAIT in case they aren't listening for S_EXIT stops or
284 * decided to wait again after we told them we are exiting.
286 p->p_flag |= P_WEXIT;
290 * Wait for any processes that have a hold on our vmspace to
291 * release their reference.
293 while (p->p_lock > 0)
294 msleep(&p->p_lock, &p->p_mtx, PWAIT, "exithold", 0);
297 /* Drain the limit callout while we don't have the proc locked */
298 callout_drain(&p->p_limco);
302 * The Sun BSM exit token contains two components: an exit status as
303 * passed to exit(), and a return value to indicate what sort of exit
304 * it was. The exit status is WEXITSTATUS(rv), but it's not clear
305 * what the return value is.
307 AUDIT_ARG_EXIT(rval, 0);
308 AUDIT_SYSCALL_EXIT(0, td);
311 /* Are we a task leader with peers? */
312 if (p->p_peers != NULL && p == p->p_leader) {
313 mtx_lock(&ppeers_lock);
317 kern_psignal(q, SIGKILL);
321 while (p->p_peers != NULL)
322 msleep(p, &ppeers_lock, PWAIT, "exit1", 0);
323 mtx_unlock(&ppeers_lock);
327 * Check if any loadable modules need anything done at process exit.
328 * E.g. SYSV IPC stuff.
329 * Event handler could change exit status.
330 * XXX what if one of these generates an error?
332 EVENTHANDLER_INVOKE(process_exit, p);
335 * If parent is waiting for us to exit or exec,
336 * P_PPWAIT is set; we will wakeup the parent below.
340 p->p_flag &= ~(P_TRACED | P_PPWAIT | P_PPTRACE);
344 * Stop the real interval timer. If the handler is currently
345 * executing, prevent it from rearming itself and let it finish.
347 if (timevalisset(&p->p_realtimer.it_value) &&
348 _callout_stop_safe(&p->p_itcallout, CS_EXECUTING, NULL) == 0) {
349 timevalclear(&p->p_realtimer.it_interval);
350 msleep(&p->p_itcallout, &p->p_mtx, PWAIT, "ritwait", 0);
351 KASSERT(!timevalisset(&p->p_realtimer.it_value),
352 ("realtime timer is still armed"));
357 umtx_thread_exit(td);
360 * Reset any sigio structures pointing to us as a result of
361 * F_SETOWN with our pid.
363 funsetownlst(&p->p_sigiolst);
366 * If this process has an nlminfo data area (for lockd), release it
368 if (nlminfo_release_p != NULL && p->p_nlminfo != NULL)
369 (*nlminfo_release_p)(p);
372 * Close open files and release open-file table.
378 * If this thread tickled GEOM, we need to wait for the giggling to
379 * stop before we return to userland
381 if (td->td_pflags & TDP_GEOM)
385 * Remove ourself from our leader's peer list and wake our leader.
387 if (p->p_leader->p_peers != NULL) {
388 mtx_lock(&ppeers_lock);
389 if (p->p_leader->p_peers != NULL) {
391 while (q->p_peers != p)
393 q->p_peers = p->p_peers;
396 mtx_unlock(&ppeers_lock);
401 (void)acct_process(td);
407 * Release reference to text vnode
409 if (p->p_textvp != NULL) {
415 * Release our limits structure.
417 lim_free(p->p_limit);
423 * Remove proc from allproc queue and pidhash chain.
424 * Place onto zombproc. Unlink from parent's child list.
426 sx_xlock(&allproc_lock);
427 LIST_REMOVE(p, p_list);
428 LIST_INSERT_HEAD(&zombproc, p, p_list);
429 LIST_REMOVE(p, p_hash);
430 sx_xunlock(&allproc_lock);
433 * Call machine-dependent code to release any
434 * machine-dependent resources other than the address space.
435 * The address space is released by "vmspace_exitfree(p)" in
440 WITNESS_WARN(WARN_PANIC, NULL, "process (pid %d) exiting", p->p_pid);
443 * Reparent all children processes:
444 * - traced ones to the original parent (or init if we are that parent)
447 sx_xlock(&proctree_lock);
448 q = LIST_FIRST(&p->p_children);
449 if (q != NULL) /* only need this if any child is S_ZOMB */
451 for (; q != NULL; q = nq) {
452 nq = LIST_NEXT(q, p_sibling);
454 q->p_sigparent = SIGCHLD;
456 if (!(q->p_flag & P_TRACED)) {
457 proc_reparent(q, q->p_reaper);
460 * Traced processes are killed since their existence
461 * means someone is screwing up.
463 t = proc_realparent(q);
465 proc_reparent(q, q->p_reaper);
472 * Since q was found on our children list, the
473 * proc_reparent() call moved q to the orphan
474 * list due to present P_TRACED flag. Clear
475 * orphan link for q now while q is locked.
478 q->p_flag &= ~(P_TRACED | P_STOPPED_TRACE);
479 q->p_flag2 &= ~P2_PTRACE_FSTP;
481 FOREACH_THREAD_IN_PROC(q, tdt) {
482 tdt->td_dbgflags &= ~(TDB_SUSPEND | TDB_XSIG |
485 kern_psignal(q, SIGKILL);
491 * Also get rid of our orphans.
493 while ((q = LIST_FIRST(&p->p_orphans)) != NULL) {
495 CTR2(KTR_PTRACE, "exit: pid %d, clearing orphan %d", p->p_pid,
501 /* Save exit status. */
505 /* Tell the prison that we are gone. */
506 prison_proc_free(p->p_ucred->cr_prison);
510 * Tell the DTrace fasttrap provider about the exit if it
511 * has declared an interest.
513 if (dtrace_fasttrap_exit)
514 dtrace_fasttrap_exit(p);
518 * Notify interested parties of our demise.
520 KNOTE_LOCKED(p->p_klist, NOTE_EXIT);
523 int reason = CLD_EXITED;
524 if (WCOREDUMP(signo))
526 else if (WIFSIGNALED(signo))
528 SDT_PROBE1(proc, , , exit, reason);
532 * If this is a process with a descriptor, we may not need to deliver
533 * a signal to the parent. proctree_lock is held over
534 * procdesc_exit() to serialize concurrent calls to close() and
537 if (p->p_procdesc == NULL || procdesc_exit(p)) {
539 * Notify parent that we're gone. If parent has the
540 * PS_NOCLDWAIT flag set, or if the handler is set to SIG_IGN,
541 * notify process 1 instead (and hope it will handle this
544 PROC_LOCK(p->p_pptr);
545 mtx_lock(&p->p_pptr->p_sigacts->ps_mtx);
546 if (p->p_pptr->p_sigacts->ps_flag &
547 (PS_NOCLDWAIT | PS_CLDSIGIGN)) {
550 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
553 proc_reparent(p, p->p_reaper);
554 p->p_sigparent = SIGCHLD;
555 PROC_LOCK(p->p_pptr);
558 * Notify parent, so in case he was wait(2)ing or
559 * executing waitpid(2) with our pid, he will
564 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
566 if (p->p_pptr == p->p_reaper || p->p_pptr == initproc)
568 else if (p->p_sigparent != 0) {
569 if (p->p_sigparent == SIGCHLD)
571 else /* LINUX thread */
572 kern_psignal(p->p_pptr, p->p_sigparent);
575 PROC_LOCK(p->p_pptr);
576 sx_xunlock(&proctree_lock);
579 * The state PRS_ZOMBIE prevents other proesses from sending
580 * signal to the process, to avoid memory leak, we free memory
581 * for signal queue at the time when the state is set.
583 sigqueue_flush(&p->p_sigqueue);
584 sigqueue_flush(&td->td_sigqueue);
587 * We have to wait until after acquiring all locks before
588 * changing p_state. We need to avoid all possible context
589 * switches (including ones from blocking on a mutex) while
590 * marked as a zombie. We also have to set the zombie state
591 * before we release the parent process' proc lock to avoid
592 * a lost wakeup. So, we first call wakeup, then we grab the
593 * sched lock, update the state, and release the parent process'
597 cv_broadcast(&p->p_pwait);
598 sched_exit(p->p_pptr, td);
600 p->p_state = PRS_ZOMBIE;
601 PROC_UNLOCK(p->p_pptr);
604 * Save our children's rusage information in our exit rusage.
607 ruadd(&p->p_ru, &p->p_rux, &p->p_stats->p_cru, &p->p_crux);
611 * Make sure the scheduler takes this thread out of its tables etc.
612 * This will also release this thread's reference to the ucred.
613 * Other thread parts to release include pcb bits and such.
619 #ifndef _SYS_SYSPROTO_H_
628 sys_abort2(struct thread *td, struct abort2_args *uap)
630 struct proc *p = td->td_proc;
636 * Do it right now so we can log either proper call of abort2(), or
637 * note, that invalid argument was passed. 512 is big enough to
638 * handle 16 arguments' descriptions with additional comments.
640 sb = sbuf_new(NULL, NULL, 512, SBUF_FIXEDLEN);
642 sbuf_printf(sb, "%s(pid %d uid %d) aborted: ",
643 p->p_comm, p->p_pid, td->td_ucred->cr_uid);
645 * Since we can't return from abort2(), send SIGKILL in cases, where
646 * abort2() was called improperly
649 /* Prevent from DoSes from user-space. */
650 if (uap->nargs < 0 || uap->nargs > 16)
652 if (uap->nargs > 0) {
653 if (uap->args == NULL)
655 error = copyin(uap->args, uargs, uap->nargs * sizeof(void *));
660 * Limit size of 'reason' string to 128. Will fit even when
661 * maximal number of arguments was chosen to be logged.
663 if (uap->why != NULL) {
664 error = sbuf_copyin(sb, uap->why, 128);
668 sbuf_printf(sb, "(null)");
670 if (uap->nargs > 0) {
671 sbuf_printf(sb, "(");
672 for (i = 0;i < uap->nargs; i++)
673 sbuf_printf(sb, "%s%p", i == 0 ? "" : ", ", uargs[i]);
674 sbuf_printf(sb, ")");
677 * Final stage: arguments were proper, string has been
678 * successfully copied from userspace, and copying pointers
679 * from user-space succeed.
683 if (sig == SIGKILL) {
685 sbuf_printf(sb, " (Reason text inaccessible)");
689 log(LOG_INFO, "%s", sbuf_data(sb));
698 * The dirty work is handled by kern_wait().
701 owait(struct thread *td, struct owait_args *uap __unused)
705 error = kern_wait(td, WAIT_ANY, &status, 0, NULL);
707 td->td_retval[1] = status;
710 #endif /* COMPAT_43 */
713 * The dirty work is handled by kern_wait().
716 sys_wait4(struct thread *td, struct wait4_args *uap)
718 struct rusage ru, *rup;
721 if (uap->rusage != NULL)
725 error = kern_wait(td, uap->pid, &status, uap->options, rup);
726 if (uap->status != NULL && error == 0)
727 error = copyout(&status, uap->status, sizeof(status));
728 if (uap->rusage != NULL && error == 0)
729 error = copyout(&ru, uap->rusage, sizeof(struct rusage));
734 sys_wait6(struct thread *td, struct wait6_args *uap)
736 struct __wrusage wru, *wrup;
742 idtype = uap->idtype;
745 if (uap->wrusage != NULL)
750 if (uap->info != NULL) {
752 bzero(sip, sizeof(*sip));
757 * We expect all callers of wait6() to know about WEXITED and
760 error = kern_wait6(td, idtype, id, &status, uap->options, wrup, sip);
762 if (uap->status != NULL && error == 0)
763 error = copyout(&status, uap->status, sizeof(status));
764 if (uap->wrusage != NULL && error == 0)
765 error = copyout(&wru, uap->wrusage, sizeof(wru));
766 if (uap->info != NULL && error == 0)
767 error = copyout(&si, uap->info, sizeof(si));
772 * Reap the remains of a zombie process and optionally return status and
773 * rusage. Asserts and will release both the proctree_lock and the process
774 * lock as part of its work.
777 proc_reap(struct thread *td, struct proc *p, int *status, int options)
781 sx_assert(&proctree_lock, SA_XLOCKED);
782 PROC_LOCK_ASSERT(p, MA_OWNED);
783 PROC_SLOCK_ASSERT(p, MA_OWNED);
784 KASSERT(p->p_state == PRS_ZOMBIE, ("proc_reap: !PRS_ZOMBIE"));
790 *status = KW_EXITCODE(p->p_xexit, p->p_xsig);
791 if (options & WNOWAIT) {
793 * Only poll, returning the status. Caller does not wish to
794 * release the proc struct just yet.
797 sx_xunlock(&proctree_lock);
802 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 && p->p_oppid != p->p_pptr->p_pid) {
811 t = proc_realparent(p);
815 "wait: traced child %d moved back to parent %d", p->p_pid,
820 pksignal(t, SIGCHLD, p->p_ksi);
822 cv_broadcast(&p->p_pwait);
824 sx_xunlock(&proctree_lock);
831 * Remove other references to this process to ensure we have an
832 * exclusive reference.
834 sx_xlock(&allproc_lock);
835 LIST_REMOVE(p, p_list); /* off zombproc */
836 sx_xunlock(&allproc_lock);
837 LIST_REMOVE(p, p_sibling);
838 reaper_abandon_children(p, true);
839 LIST_REMOVE(p, p_reapsibling);
844 if (p->p_procdesc != NULL)
846 sx_xunlock(&proctree_lock);
849 knlist_detach(p->p_klist);
854 * Removal from allproc list and process group list paired with
855 * PROC_LOCK which was executed during that time should guarantee
856 * nothing can reach this process anymore. As such further locking
859 p->p_xexit = p->p_xsig = 0; /* XXX: why? */
862 ruadd(&q->p_stats->p_cru, &q->p_crux, &p->p_ru, &p->p_rux);
866 * Decrement the count of procs running with this uid.
868 (void)chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0);
871 * Destroy resource accounting information associated with the process.
876 racct_sub(p, RACCT_NPROC, 1);
883 * Free credentials, arguments, and sigacts.
886 proc_set_cred(p, NULL);
887 pargs_drop(p->p_args);
889 sigacts_free(p->p_sigacts);
893 * Do any thread-system specific cleanups.
898 * Give vm and machine-dependent layer a chance to free anything that
899 * cpu_exit couldn't release while still running in process context.
906 * Free any domain policy that's still hiding around.
908 vm_domain_policy_cleanup(&p->p_vm_dom_policy);
910 KASSERT(FIRST_THREAD_IN_PROC(p),
911 ("proc_reap: no residual thread!"));
912 uma_zfree(proc_zone, p);
913 atomic_add_int(&nprocs, -1);
917 proc_to_reap(struct thread *td, struct proc *p, idtype_t idtype, id_t id,
918 int *status, int options, struct __wrusage *wrusage, siginfo_t *siginfo,
923 sx_assert(&proctree_lock, SA_XLOCKED);
929 if (p->p_procdesc != NULL) {
935 if (p->p_pid != (pid_t)id) {
941 if (p->p_pgid != (pid_t)id) {
947 if (p->p_session->s_sid != (pid_t)id) {
953 if (p->p_ucred->cr_uid != (uid_t)id) {
959 if (p->p_ucred->cr_gid != (gid_t)id) {
965 if (p->p_ucred->cr_prison->pr_id != (int)id) {
971 * It seems that the thread structures get zeroed out
972 * at process exit. This makes it impossible to
973 * support P_SETID, P_CID or P_CPUID.
980 if (p_canwait(td, p)) {
985 if (((options & WEXITED) == 0) && (p->p_state == PRS_ZOMBIE)) {
991 * This special case handles a kthread spawned by linux_clone
992 * (see linux_misc.c). The linux_wait4 and linux_waitpid
993 * functions need to be able to distinguish between waiting
994 * on a process and waiting on a thread. It is a thread if
995 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option
996 * signifies we want to wait for threads and not processes.
998 if ((p->p_sigparent != SIGCHLD) ^
999 ((options & WLINUXCLONE) != 0)) {
1004 if (siginfo != NULL) {
1005 bzero(siginfo, sizeof(*siginfo));
1006 siginfo->si_errno = 0;
1009 * SUSv4 requires that the si_signo value is always
1010 * SIGCHLD. Obey it despite the rfork(2) interface
1011 * allows to request other signal for child exit
1014 siginfo->si_signo = SIGCHLD;
1017 * This is still a rough estimate. We will fix the
1018 * cases TRAPPED, STOPPED, and CONTINUED later.
1020 if (WCOREDUMP(p->p_xsig)) {
1021 siginfo->si_code = CLD_DUMPED;
1022 siginfo->si_status = WTERMSIG(p->p_xsig);
1023 } else if (WIFSIGNALED(p->p_xsig)) {
1024 siginfo->si_code = CLD_KILLED;
1025 siginfo->si_status = WTERMSIG(p->p_xsig);
1027 siginfo->si_code = CLD_EXITED;
1028 siginfo->si_status = p->p_xexit;
1031 siginfo->si_pid = p->p_pid;
1032 siginfo->si_uid = p->p_ucred->cr_uid;
1035 * The si_addr field would be useful additional
1036 * detail, but apparently the PC value may be lost
1037 * when we reach this point. bzero() above sets
1038 * siginfo->si_addr to NULL.
1043 * There should be no reason to limit resources usage info to
1044 * exited processes only. A snapshot about any resources used
1045 * by a stopped process may be exactly what is needed.
1047 if (wrusage != NULL) {
1048 rup = &wrusage->wru_self;
1051 calcru(p, &rup->ru_utime, &rup->ru_stime);
1054 rup = &wrusage->wru_children;
1055 *rup = p->p_stats->p_cru;
1056 calccru(p, &rup->ru_utime, &rup->ru_stime);
1059 if (p->p_state == PRS_ZOMBIE && !check_only) {
1061 proc_reap(td, p, status, options);
1069 kern_wait(struct thread *td, pid_t pid, int *status, int options,
1070 struct rusage *rusage)
1072 struct __wrusage wru, *wrup;
1078 * Translate the special pid values into the (idtype, pid)
1079 * pair for kern_wait6. The WAIT_MYPGRP case is handled by
1080 * kern_wait6() on its own.
1082 if (pid == WAIT_ANY) {
1085 } else if (pid < 0) {
1099 * For backward compatibility we implicitly add flags WEXITED
1100 * and WTRAPPED here.
1102 options |= WEXITED | WTRAPPED;
1103 ret = kern_wait6(td, idtype, id, status, options, wrup, NULL);
1105 *rusage = wru.wru_self;
1110 kern_wait6(struct thread *td, idtype_t idtype, id_t id, int *status,
1111 int options, struct __wrusage *wrusage, siginfo_t *siginfo)
1115 int error, nfound, ret;
1117 AUDIT_ARG_VALUE((int)idtype); /* XXX - This is likely wrong! */
1118 AUDIT_ARG_PID((pid_t)id); /* XXX - This may be wrong! */
1119 AUDIT_ARG_VALUE(options);
1123 if ((pid_t)id == WAIT_MYPGRP && (idtype == P_PID || idtype == P_PGID)) {
1125 id = (id_t)q->p_pgid;
1130 /* If we don't know the option, just return. */
1131 if ((options & ~(WUNTRACED | WNOHANG | WCONTINUED | WNOWAIT |
1132 WEXITED | WTRAPPED | WLINUXCLONE)) != 0)
1134 if ((options & (WEXITED | WUNTRACED | WCONTINUED | WTRAPPED)) == 0) {
1136 * We will be unable to find any matching processes,
1137 * because there are no known events to look for.
1138 * Prefer to return error instead of blocking
1145 if (q->p_flag & P_STATCHILD) {
1147 q->p_flag &= ~P_STATCHILD;
1151 sx_xlock(&proctree_lock);
1152 LIST_FOREACH(p, &q->p_children, p_sibling) {
1154 ret = proc_to_reap(td, p, idtype, id, status, options,
1155 wrusage, siginfo, 0);
1161 td->td_retval[0] = pid;
1168 if ((options & WTRAPPED) != 0 &&
1169 (p->p_flag & P_TRACED) != 0 &&
1170 (p->p_flag & (P_STOPPED_TRACE | P_STOPPED_SIG)) != 0 &&
1171 (p->p_suspcount == p->p_numthreads) &&
1172 ((p->p_flag & P_WAITED) == 0)) {
1174 if ((options & WNOWAIT) == 0)
1175 p->p_flag |= P_WAITED;
1176 sx_xunlock(&proctree_lock);
1179 *status = W_STOPCODE(p->p_xsig);
1180 if (siginfo != NULL) {
1181 siginfo->si_status = p->p_xsig;
1182 siginfo->si_code = CLD_TRAPPED;
1184 if ((options & WNOWAIT) == 0) {
1186 sigqueue_take(p->p_ksi);
1191 "wait: returning trapped pid %d status %#x (xstat %d) xthread %d",
1192 p->p_pid, W_STOPCODE(p->p_xsig), p->p_xsig,
1193 p->p_xthread != NULL ? p->p_xthread->td_tid : -1);
1195 td->td_retval[0] = pid;
1198 if ((options & WUNTRACED) != 0 &&
1199 (p->p_flag & P_STOPPED_SIG) != 0 &&
1200 (p->p_suspcount == p->p_numthreads) &&
1201 ((p->p_flag & P_WAITED) == 0)) {
1203 if ((options & WNOWAIT) == 0)
1204 p->p_flag |= P_WAITED;
1205 sx_xunlock(&proctree_lock);
1208 *status = W_STOPCODE(p->p_xsig);
1209 if (siginfo != NULL) {
1210 siginfo->si_status = p->p_xsig;
1211 siginfo->si_code = CLD_STOPPED;
1213 if ((options & WNOWAIT) == 0) {
1215 sigqueue_take(p->p_ksi);
1220 td->td_retval[0] = pid;
1224 if ((options & WCONTINUED) != 0 &&
1225 (p->p_flag & P_CONTINUED) != 0) {
1226 sx_xunlock(&proctree_lock);
1227 if ((options & WNOWAIT) == 0) {
1228 p->p_flag &= ~P_CONTINUED;
1230 sigqueue_take(p->p_ksi);
1237 if (siginfo != NULL) {
1238 siginfo->si_status = SIGCONT;
1239 siginfo->si_code = CLD_CONTINUED;
1241 td->td_retval[0] = pid;
1248 * Look in the orphans list too, to allow the parent to
1249 * collect it's child exit status even if child is being
1252 * Debugger detaches from the parent upon successful
1253 * switch-over from parent to child. At this point due to
1254 * re-parenting the parent loses the child to debugger and a
1255 * wait4(2) call would report that it has no children to wait
1256 * for. By maintaining a list of orphans we allow the parent
1257 * to successfully wait until the child becomes a zombie.
1260 LIST_FOREACH(p, &q->p_orphans, p_orphan) {
1261 ret = proc_to_reap(td, p, idtype, id, NULL, options,
1264 KASSERT(ret != -1, ("reaped an orphan (pid %d)",
1265 (int)td->td_retval[0]));
1272 sx_xunlock(&proctree_lock);
1275 if (options & WNOHANG) {
1276 sx_xunlock(&proctree_lock);
1277 td->td_retval[0] = 0;
1281 sx_xunlock(&proctree_lock);
1282 if (q->p_flag & P_STATCHILD) {
1283 q->p_flag &= ~P_STATCHILD;
1286 error = msleep(q, &q->p_mtx, PWAIT | PCATCH, "wait", 0);
1294 * Make process 'parent' the new parent of process 'child'.
1295 * Must be called with an exclusive hold of proctree lock.
1298 proc_reparent(struct proc *child, struct proc *parent)
1301 sx_assert(&proctree_lock, SX_XLOCKED);
1302 PROC_LOCK_ASSERT(child, MA_OWNED);
1303 if (child->p_pptr == parent)
1306 PROC_LOCK(child->p_pptr);
1307 sigqueue_take(child->p_ksi);
1308 PROC_UNLOCK(child->p_pptr);
1309 LIST_REMOVE(child, p_sibling);
1310 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
1312 clear_orphan(child);
1313 if (child->p_flag & P_TRACED) {
1314 if (LIST_EMPTY(&child->p_pptr->p_orphans)) {
1315 child->p_treeflag |= P_TREE_FIRST_ORPHAN;
1316 LIST_INSERT_HEAD(&child->p_pptr->p_orphans, child,
1319 LIST_INSERT_AFTER(LIST_FIRST(&child->p_pptr->p_orphans),
1322 child->p_treeflag |= P_TREE_ORPHANED;
1325 child->p_pptr = parent;