2 * Copyright (c) 1982, 1986, 1989, 1991, 1993
3 * The Regents of the University of California. All rights reserved.
4 * (c) UNIX System Laboratories, Inc.
5 * All or some portions of this file are derived from material licensed
6 * to the University of California by American Telephone and Telegraph
7 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8 * the permission of UNIX System Laboratories, Inc.
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 4. Neither the name of the University nor the names of its contributors
19 * may be used to endorse or promote products derived from this software
20 * without specific prior written permission.
22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * @(#)kern_exit.c 8.7 (Berkeley) 2/12/94
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD$");
40 #include "opt_compat.h"
41 #include "opt_ktrace.h"
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/sysproto.h>
46 #include <sys/capsicum.h>
47 #include <sys/eventhandler.h>
48 #include <sys/kernel.h>
49 #include <sys/malloc.h>
51 #include <sys/mutex.h>
53 #include <sys/procdesc.h>
54 #include <sys/pioctl.h>
58 #include <sys/vmmeter.h>
59 #include <sys/vnode.h>
60 #include <sys/racct.h>
61 #include <sys/resourcevar.h>
63 #include <sys/signalvar.h>
64 #include <sys/sched.h>
66 #include <sys/syscallsubr.h>
67 #include <sys/syslog.h>
68 #include <sys/ptrace.h>
69 #include <sys/acct.h> /* for acct_process() function prototype */
70 #include <sys/filedesc.h>
75 #include <sys/ktrace.h>
78 #include <security/audit/audit.h>
79 #include <security/mac/mac_framework.h>
82 #include <vm/vm_extern.h>
83 #include <vm/vm_param.h>
85 #include <vm/vm_map.h>
86 #include <vm/vm_page.h>
90 #include <sys/dtrace_bsd.h>
91 dtrace_execexit_func_t dtrace_fasttrap_exit;
94 SDT_PROVIDER_DECLARE(proc);
95 SDT_PROBE_DEFINE1(proc, kernel, , exit, "int");
97 /* Hook for NFS teardown procedure. */
98 void (*nlminfo_release_p)(struct proc *p);
101 proc_realparent(struct proc *child)
103 struct proc *p, *parent;
105 sx_assert(&proctree_lock, SX_LOCKED);
106 if ((child->p_treeflag & P_TREE_ORPHANED) == 0) {
107 if (child->p_oppid == 0 ||
108 child->p_pptr->p_pid == child->p_oppid)
109 parent = child->p_pptr;
114 for (p = child; (p->p_treeflag & P_TREE_FIRST_ORPHAN) == 0;) {
115 /* Cannot use LIST_PREV(), since the list head is not known. */
116 p = __containerof(p->p_orphan.le_prev, struct proc,
118 KASSERT((p->p_treeflag & P_TREE_ORPHANED) != 0,
119 ("missing P_ORPHAN %p", p));
121 parent = __containerof(p->p_orphan.le_prev, struct proc,
127 reaper_abandon_children(struct proc *p, bool exiting)
129 struct proc *p1, *p2, *ptmp;
131 sx_assert(&proctree_lock, SX_LOCKED);
132 KASSERT(p != initproc, ("reaper_abandon_children for initproc"));
133 if ((p->p_treeflag & P_TREE_REAPER) == 0)
136 LIST_FOREACH_SAFE(p2, &p->p_reaplist, p_reapsibling, ptmp) {
137 LIST_REMOVE(p2, p_reapsibling);
139 p2->p_reapsubtree = p->p_reapsubtree;
140 LIST_INSERT_HEAD(&p1->p_reaplist, p2, p_reapsibling);
141 if (exiting && p2->p_pptr == p) {
143 proc_reparent(p2, p1);
147 KASSERT(LIST_EMPTY(&p->p_reaplist), ("p_reaplist not empty"));
148 p->p_treeflag &= ~P_TREE_REAPER;
152 clear_orphan(struct proc *p)
156 sx_assert(&proctree_lock, SA_XLOCKED);
157 if ((p->p_treeflag & P_TREE_ORPHANED) == 0)
159 if ((p->p_treeflag & P_TREE_FIRST_ORPHAN) != 0) {
160 p1 = LIST_NEXT(p, p_orphan);
162 p1->p_treeflag |= P_TREE_FIRST_ORPHAN;
163 p->p_treeflag &= ~P_TREE_FIRST_ORPHAN;
165 LIST_REMOVE(p, p_orphan);
166 p->p_treeflag &= ~P_TREE_ORPHANED;
170 * exit -- death of process.
173 sys_sys_exit(struct thread *td, struct sys_exit_args *uap)
176 exit1(td, W_EXITCODE(uap->rval, 0));
181 * Exit: deallocate address space and other resources, change proc state to
182 * zombie, and unlink proc from allproc and parent's lists. Save exit status
183 * and rusage for wait(). Check for child processes and orphan them.
186 exit1(struct thread *td, int rv)
188 struct proc *p, *nq, *q, *t;
190 struct vnode *ttyvp = NULL;
192 mtx_assert(&Giant, MA_NOTOWNED);
196 * XXX in case we're rebooting we just let init die in order to
197 * work around an unsolved stack overflow seen very late during
198 * shutdown on sparc64 when the gmirror worker process exists.
200 if (p == initproc && rebooting == 0) {
201 printf("init died (signal %d, exit %d)\n",
202 WTERMSIG(rv), WEXITSTATUS(rv));
203 panic("Going nowhere without my init!");
207 * MUST abort all other threads before proceeding past here.
211 * First check if some other thread or external request got
212 * here before us. If so, act appropriately: exit or suspend.
213 * We must ensure that stop requests are handled before we set
216 thread_suspend_check(0);
217 while (p->p_flag & P_HADTHREADS) {
219 * Kill off the other threads. This requires
220 * some co-operation from other parts of the kernel
221 * so it may not be instantaneous. With this state set
222 * any thread entering the kernel from userspace will
223 * thread_exit() in trap(). Any thread attempting to
224 * sleep will return immediately with EINTR or EWOULDBLOCK
225 * which will hopefully force them to back out to userland
226 * freeing resources as they go. Any thread attempting
227 * to return to userland will thread_exit() from userret().
228 * thread_exit() will unsuspend us when the last of the
229 * other threads exits.
230 * If there is already a thread singler after resumption,
231 * calling thread_single will fail; in that case, we just
232 * re-check all suspension request, the thread should
233 * either be suspended there or exit.
235 if (!thread_single(p, SINGLE_EXIT))
237 * All other activity in this process is now
238 * stopped. Threading support has been turned
243 * Recheck for new stop or suspend requests which
244 * might appear while process lock was dropped in
247 thread_suspend_check(0);
249 KASSERT(p->p_numthreads == 1,
250 ("exit1: proc %p exiting with %d threads", p, p->p_numthreads));
251 racct_sub(p, RACCT_NTHR, 1);
253 * Wakeup anyone in procfs' PIOCWAIT. They should have a hold
254 * on our vmspace, so we should block below until they have
255 * released their reference to us. Note that if they have
256 * requested S_EXIT stops we will block here until they ack
259 _STOPEVENT(p, S_EXIT, rv);
262 * Ignore any pending request to stop due to a stop signal.
263 * Once P_WEXIT is set, future requests will be ignored as
266 p->p_flag &= ~P_STOPPED_SIG;
267 KASSERT(!P_SHOULDSTOP(p), ("exiting process is stopped"));
270 * Note that we are exiting and do another wakeup of anyone in
271 * PIOCWAIT in case they aren't listening for S_EXIT stops or
272 * decided to wait again after we told them we are exiting.
274 p->p_flag |= P_WEXIT;
278 * Wait for any processes that have a hold on our vmspace to
279 * release their reference.
281 while (p->p_lock > 0)
282 msleep(&p->p_lock, &p->p_mtx, PWAIT, "exithold", 0);
284 p->p_xstat = rv; /* Let event handler change exit status */
286 /* Drain the limit callout while we don't have the proc locked */
287 callout_drain(&p->p_limco);
291 * The Sun BSM exit token contains two components: an exit status as
292 * passed to exit(), and a return value to indicate what sort of exit
293 * it was. The exit status is WEXITSTATUS(rv), but it's not clear
294 * what the return value is.
296 AUDIT_ARG_EXIT(WEXITSTATUS(rv), 0);
297 AUDIT_SYSCALL_EXIT(0, td);
300 /* Are we a task leader with peers? */
301 if (p->p_peers != NULL && p == p->p_leader) {
302 mtx_lock(&ppeers_lock);
306 kern_psignal(q, SIGKILL);
310 while (p->p_peers != NULL)
311 msleep(p, &ppeers_lock, PWAIT, "exit1", 0);
312 mtx_unlock(&ppeers_lock);
316 * Check if any loadable modules need anything done at process exit.
317 * E.g. SYSV IPC stuff
318 * XXX what if one of these generates an error?
320 EVENTHANDLER_INVOKE(process_exit, p);
323 * If parent is waiting for us to exit or exec,
324 * P_PPWAIT is set; we will wakeup the parent below.
327 rv = p->p_xstat; /* Event handler could change exit status */
329 p->p_flag &= ~(P_TRACED | P_PPWAIT | P_PPTRACE);
332 * Stop the real interval timer. If the handler is currently
333 * executing, prevent it from rearming itself and let it finish.
335 if (timevalisset(&p->p_realtimer.it_value) &&
336 callout_stop(&p->p_itcallout) == 0) {
337 timevalclear(&p->p_realtimer.it_interval);
338 msleep(&p->p_itcallout, &p->p_mtx, PWAIT, "ritwait", 0);
339 KASSERT(!timevalisset(&p->p_realtimer.it_value),
340 ("realtime timer is still armed"));
345 * Reset any sigio structures pointing to us as a result of
346 * F_SETOWN with our pid.
348 funsetownlst(&p->p_sigiolst);
351 * If this process has an nlminfo data area (for lockd), release it
353 if (nlminfo_release_p != NULL && p->p_nlminfo != NULL)
354 (*nlminfo_release_p)(p);
357 * Close open files and release open-file table.
363 * If this thread tickled GEOM, we need to wait for the giggling to
364 * stop before we return to userland
366 if (td->td_pflags & TDP_GEOM)
370 * Remove ourself from our leader's peer list and wake our leader.
372 if (p->p_leader->p_peers != NULL) {
373 mtx_lock(&ppeers_lock);
374 if (p->p_leader->p_peers != NULL) {
376 while (q->p_peers != p)
378 q->p_peers = p->p_peers;
381 mtx_unlock(&ppeers_lock);
386 sx_xlock(&proctree_lock);
387 if (SESS_LEADER(p)) {
388 struct session *sp = p->p_session;
392 * s_ttyp is not zero'd; we use this to indicate that
393 * the session once had a controlling terminal. (for
394 * logging and informational purposes)
405 * Signal foreground pgrp and revoke access to
406 * controlling terminal if it has not been revoked
409 * Because the TTY may have been revoked in the mean
410 * time and could already have a new session associated
411 * with it, make sure we don't send a SIGHUP to a
412 * foreground process group that does not belong to this
418 if (tp->t_session == sp)
419 tty_signal_pgrp(tp, SIGHUP);
424 sx_xunlock(&proctree_lock);
425 if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
426 VOP_REVOKE(ttyvp, REVOKEALL);
427 VOP_UNLOCK(ttyvp, 0);
429 sx_xlock(&proctree_lock);
432 fixjobc(p, p->p_pgrp, 0);
433 sx_xunlock(&proctree_lock);
434 (void)acct_process(td);
436 /* Release the TTY now we've unlocked everything. */
443 * Release reference to text vnode
445 if (p->p_textvp != NULL) {
451 * Release our limits structure.
453 lim_free(p->p_limit);
459 * Remove proc from allproc queue and pidhash chain.
460 * Place onto zombproc. Unlink from parent's child list.
462 sx_xlock(&allproc_lock);
463 LIST_REMOVE(p, p_list);
464 LIST_INSERT_HEAD(&zombproc, p, p_list);
465 LIST_REMOVE(p, p_hash);
466 sx_xunlock(&allproc_lock);
469 * Call machine-dependent code to release any
470 * machine-dependent resources other than the address space.
471 * The address space is released by "vmspace_exitfree(p)" in
476 WITNESS_WARN(WARN_PANIC, NULL, "process (pid %d) exiting", p->p_pid);
479 * Reparent all children processes:
480 * - traced ones to the original parent (or init if we are that parent)
483 sx_xlock(&proctree_lock);
484 q = LIST_FIRST(&p->p_children);
485 if (q != NULL) /* only need this if any child is S_ZOMB */
487 for (; q != NULL; q = nq) {
488 nq = LIST_NEXT(q, p_sibling);
490 q->p_sigparent = SIGCHLD;
492 if (!(q->p_flag & P_TRACED)) {
493 proc_reparent(q, q->p_reaper);
496 * Traced processes are killed since their existence
497 * means someone is screwing up.
499 t = proc_realparent(q);
501 proc_reparent(q, q->p_reaper);
508 * Since q was found on our children list, the
509 * proc_reparent() call moved q to the orphan
510 * list due to present P_TRACED flag. Clear
511 * orphan link for q now while q is locked.
514 q->p_flag &= ~(P_TRACED | P_STOPPED_TRACE);
515 FOREACH_THREAD_IN_PROC(q, tdt)
516 tdt->td_dbgflags &= ~TDB_SUSPEND;
517 kern_psignal(q, SIGKILL);
523 * Also get rid of our orphans.
525 while ((q = LIST_FIRST(&p->p_orphans)) != NULL) {
531 /* Save exit status. */
535 /* Tell the prison that we are gone. */
536 prison_proc_free(p->p_ucred->cr_prison);
540 * Tell the DTrace fasttrap provider about the exit if it
541 * has declared an interest.
543 if (dtrace_fasttrap_exit)
544 dtrace_fasttrap_exit(p);
548 * Notify interested parties of our demise.
550 KNOTE_LOCKED(&p->p_klist, NOTE_EXIT);
553 int reason = CLD_EXITED;
556 else if (WIFSIGNALED(rv))
558 SDT_PROBE(proc, kernel, , exit, reason, 0, 0, 0, 0);
562 * Just delete all entries in the p_klist. At this point we won't
563 * report any more events, and there are nasty race conditions that
564 * can beat us if we don't.
566 knlist_clear(&p->p_klist, 1);
569 * If this is a process with a descriptor, we may not need to deliver
570 * a signal to the parent. proctree_lock is held over
571 * procdesc_exit() to serialize concurrent calls to close() and
574 if (p->p_procdesc == NULL || procdesc_exit(p)) {
576 * Notify parent that we're gone. If parent has the
577 * PS_NOCLDWAIT flag set, or if the handler is set to SIG_IGN,
578 * notify process 1 instead (and hope it will handle this
581 PROC_LOCK(p->p_pptr);
582 mtx_lock(&p->p_pptr->p_sigacts->ps_mtx);
583 if (p->p_pptr->p_sigacts->ps_flag &
584 (PS_NOCLDWAIT | PS_CLDSIGIGN)) {
587 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
590 proc_reparent(p, p->p_reaper);
591 p->p_sigparent = SIGCHLD;
592 PROC_LOCK(p->p_pptr);
595 * Notify parent, so in case he was wait(2)ing or
596 * executing waitpid(2) with our pid, he will
601 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
603 if (p->p_pptr == p->p_reaper || p->p_pptr == initproc)
605 else if (p->p_sigparent != 0) {
606 if (p->p_sigparent == SIGCHLD)
608 else /* LINUX thread */
609 kern_psignal(p->p_pptr, p->p_sigparent);
612 PROC_LOCK(p->p_pptr);
613 sx_xunlock(&proctree_lock);
616 * The state PRS_ZOMBIE prevents other proesses from sending
617 * signal to the process, to avoid memory leak, we free memory
618 * for signal queue at the time when the state is set.
620 sigqueue_flush(&p->p_sigqueue);
621 sigqueue_flush(&td->td_sigqueue);
624 * We have to wait until after acquiring all locks before
625 * changing p_state. We need to avoid all possible context
626 * switches (including ones from blocking on a mutex) while
627 * marked as a zombie. We also have to set the zombie state
628 * before we release the parent process' proc lock to avoid
629 * a lost wakeup. So, we first call wakeup, then we grab the
630 * sched lock, update the state, and release the parent process'
634 cv_broadcast(&p->p_pwait);
635 sched_exit(p->p_pptr, td);
637 p->p_state = PRS_ZOMBIE;
638 PROC_UNLOCK(p->p_pptr);
641 * Hopefully no one will try to deliver a signal to the process this
644 knlist_destroy(&p->p_klist);
647 * Save our children's rusage information in our exit rusage.
650 ruadd(&p->p_ru, &p->p_rux, &p->p_stats->p_cru, &p->p_crux);
654 * Make sure the scheduler takes this thread out of its tables etc.
655 * This will also release this thread's reference to the ucred.
656 * Other thread parts to release include pcb bits and such.
662 #ifndef _SYS_SYSPROTO_H_
671 sys_abort2(struct thread *td, struct abort2_args *uap)
673 struct proc *p = td->td_proc;
679 * Do it right now so we can log either proper call of abort2(), or
680 * note, that invalid argument was passed. 512 is big enough to
681 * handle 16 arguments' descriptions with additional comments.
683 sb = sbuf_new(NULL, NULL, 512, SBUF_FIXEDLEN);
685 sbuf_printf(sb, "%s(pid %d uid %d) aborted: ",
686 p->p_comm, p->p_pid, td->td_ucred->cr_uid);
688 * Since we can't return from abort2(), send SIGKILL in cases, where
689 * abort2() was called improperly
692 /* Prevent from DoSes from user-space. */
693 if (uap->nargs < 0 || uap->nargs > 16)
695 if (uap->nargs > 0) {
696 if (uap->args == NULL)
698 error = copyin(uap->args, uargs, uap->nargs * sizeof(void *));
703 * Limit size of 'reason' string to 128. Will fit even when
704 * maximal number of arguments was chosen to be logged.
706 if (uap->why != NULL) {
707 error = sbuf_copyin(sb, uap->why, 128);
711 sbuf_printf(sb, "(null)");
713 if (uap->nargs > 0) {
714 sbuf_printf(sb, "(");
715 for (i = 0;i < uap->nargs; i++)
716 sbuf_printf(sb, "%s%p", i == 0 ? "" : ", ", uargs[i]);
717 sbuf_printf(sb, ")");
720 * Final stage: arguments were proper, string has been
721 * successfully copied from userspace, and copying pointers
722 * from user-space succeed.
726 if (sig == SIGKILL) {
728 sbuf_printf(sb, " (Reason text inaccessible)");
732 log(LOG_INFO, "%s", sbuf_data(sb));
734 exit1(td, W_EXITCODE(0, sig));
741 * The dirty work is handled by kern_wait().
744 owait(struct thread *td, struct owait_args *uap __unused)
748 error = kern_wait(td, WAIT_ANY, &status, 0, NULL);
750 td->td_retval[1] = status;
753 #endif /* COMPAT_43 */
756 * The dirty work is handled by kern_wait().
759 sys_wait4(struct thread *td, struct wait4_args *uap)
761 struct rusage ru, *rup;
764 if (uap->rusage != NULL)
768 error = kern_wait(td, uap->pid, &status, uap->options, rup);
769 if (uap->status != NULL && error == 0)
770 error = copyout(&status, uap->status, sizeof(status));
771 if (uap->rusage != NULL && error == 0)
772 error = copyout(&ru, uap->rusage, sizeof(struct rusage));
777 sys_wait6(struct thread *td, struct wait6_args *uap)
779 struct __wrusage wru, *wrup;
785 idtype = uap->idtype;
788 if (uap->wrusage != NULL)
793 if (uap->info != NULL) {
795 bzero(sip, sizeof(*sip));
800 * We expect all callers of wait6() to know about WEXITED and
803 error = kern_wait6(td, idtype, id, &status, uap->options, wrup, sip);
805 if (uap->status != NULL && error == 0)
806 error = copyout(&status, uap->status, sizeof(status));
807 if (uap->wrusage != NULL && error == 0)
808 error = copyout(&wru, uap->wrusage, sizeof(wru));
809 if (uap->info != NULL && error == 0)
810 error = copyout(&si, uap->info, sizeof(si));
815 * Reap the remains of a zombie process and optionally return status and
816 * rusage. Asserts and will release both the proctree_lock and the process
817 * lock as part of its work.
820 proc_reap(struct thread *td, struct proc *p, int *status, int options)
824 sx_assert(&proctree_lock, SA_XLOCKED);
825 PROC_LOCK_ASSERT(p, MA_OWNED);
826 PROC_SLOCK_ASSERT(p, MA_OWNED);
827 KASSERT(p->p_state == PRS_ZOMBIE, ("proc_reap: !PRS_ZOMBIE"));
832 td->td_retval[0] = p->p_pid;
834 *status = p->p_xstat; /* convert to int */
835 if (options & WNOWAIT) {
837 * Only poll, returning the status. Caller does not wish to
838 * release the proc struct just yet.
841 sx_xunlock(&proctree_lock);
846 sigqueue_take(p->p_ksi);
851 * If we got the child via a ptrace 'attach', we need to give it back
854 if (p->p_oppid != 0) {
855 t = proc_realparent(p);
861 pksignal(t, SIGCHLD, p->p_ksi);
863 cv_broadcast(&p->p_pwait);
865 sx_xunlock(&proctree_lock);
870 * Remove other references to this process to ensure we have an
871 * exclusive reference.
873 sx_xlock(&allproc_lock);
874 LIST_REMOVE(p, p_list); /* off zombproc */
875 sx_xunlock(&allproc_lock);
876 LIST_REMOVE(p, p_sibling);
877 reaper_abandon_children(p, true);
878 LIST_REMOVE(p, p_reapsibling);
883 if (p->p_procdesc != NULL)
885 sx_xunlock(&proctree_lock);
888 * As a side effect of this lock, we know that all other writes to
889 * this proc are visible now, so no more locking is needed for p.
892 p->p_xstat = 0; /* XXX: why? */
895 ruadd(&q->p_stats->p_cru, &q->p_crux, &p->p_ru, &p->p_rux);
899 * Decrement the count of procs running with this uid.
901 (void)chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0);
904 * Destroy resource accounting information associated with the process.
908 racct_sub(p, RACCT_NPROC, 1);
914 * Free credentials, arguments, and sigacts.
918 pargs_drop(p->p_args);
920 sigacts_free(p->p_sigacts);
924 * Do any thread-system specific cleanups.
929 * Give vm and machine-dependent layer a chance to free anything that
930 * cpu_exit couldn't release while still running in process context.
936 KASSERT(FIRST_THREAD_IN_PROC(p),
937 ("proc_reap: no residual thread!"));
938 uma_zfree(proc_zone, p);
939 sx_xlock(&allproc_lock);
941 sx_xunlock(&allproc_lock);
945 proc_to_reap(struct thread *td, struct proc *p, idtype_t idtype, id_t id,
946 int *status, int options, struct __wrusage *wrusage, siginfo_t *siginfo)
951 sx_assert(&proctree_lock, SA_XLOCKED);
960 if (p->p_pid != (pid_t)id) {
966 if (p->p_pgid != (pid_t)id) {
972 if (p->p_session->s_sid != (pid_t)id) {
978 if (p->p_ucred->cr_uid != (uid_t)id) {
984 if (p->p_ucred->cr_gid != (gid_t)id) {
990 if (p->p_ucred->cr_prison->pr_id != (int)id) {
996 * It seems that the thread structures get zeroed out
997 * at process exit. This makes it impossible to
998 * support P_SETID, P_CID or P_CPUID.
1005 if (p_canwait(td, p)) {
1010 if (((options & WEXITED) == 0) && (p->p_state == PRS_ZOMBIE)) {
1016 * This special case handles a kthread spawned by linux_clone
1017 * (see linux_misc.c). The linux_wait4 and linux_waitpid
1018 * functions need to be able to distinguish between waiting
1019 * on a process and waiting on a thread. It is a thread if
1020 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option
1021 * signifies we want to wait for threads and not processes.
1023 if ((p->p_sigparent != SIGCHLD) ^
1024 ((options & WLINUXCLONE) != 0)) {
1029 if (siginfo != NULL) {
1030 bzero(siginfo, sizeof(*siginfo));
1031 siginfo->si_errno = 0;
1034 * SUSv4 requires that the si_signo value is always
1035 * SIGCHLD. Obey it despite the rfork(2) interface
1036 * allows to request other signal for child exit
1039 siginfo->si_signo = SIGCHLD;
1042 * This is still a rough estimate. We will fix the
1043 * cases TRAPPED, STOPPED, and CONTINUED later.
1045 if (WCOREDUMP(p->p_xstat)) {
1046 siginfo->si_code = CLD_DUMPED;
1047 siginfo->si_status = WTERMSIG(p->p_xstat);
1048 } else if (WIFSIGNALED(p->p_xstat)) {
1049 siginfo->si_code = CLD_KILLED;
1050 siginfo->si_status = WTERMSIG(p->p_xstat);
1052 siginfo->si_code = CLD_EXITED;
1053 siginfo->si_status = WEXITSTATUS(p->p_xstat);
1056 siginfo->si_pid = p->p_pid;
1057 siginfo->si_uid = p->p_ucred->cr_uid;
1060 * The si_addr field would be useful additional
1061 * detail, but apparently the PC value may be lost
1062 * when we reach this point. bzero() above sets
1063 * siginfo->si_addr to NULL.
1068 * There should be no reason to limit resources usage info to
1069 * exited processes only. A snapshot about any resources used
1070 * by a stopped process may be exactly what is needed.
1072 if (wrusage != NULL) {
1073 rup = &wrusage->wru_self;
1076 calcru(p, &rup->ru_utime, &rup->ru_stime);
1079 rup = &wrusage->wru_children;
1080 *rup = p->p_stats->p_cru;
1081 calccru(p, &rup->ru_utime, &rup->ru_stime);
1084 if (p->p_state == PRS_ZOMBIE) {
1086 proc_reap(td, p, status, options);
1094 kern_wait(struct thread *td, pid_t pid, int *status, int options,
1095 struct rusage *rusage)
1097 struct __wrusage wru, *wrup;
1103 * Translate the special pid values into the (idtype, pid)
1104 * pair for kern_wait6. The WAIT_MYPGRP case is handled by
1105 * kern_wait6() on its own.
1107 if (pid == WAIT_ANY) {
1110 } else if (pid < 0) {
1124 * For backward compatibility we implicitly add flags WEXITED
1125 * and WTRAPPED here.
1127 options |= WEXITED | WTRAPPED;
1128 ret = kern_wait6(td, idtype, id, status, options, wrup, NULL);
1130 *rusage = wru.wru_self;
1135 kern_wait6(struct thread *td, idtype_t idtype, id_t id, int *status,
1136 int options, struct __wrusage *wrusage, siginfo_t *siginfo)
1139 int error, nfound, ret;
1141 AUDIT_ARG_VALUE((int)idtype); /* XXX - This is likely wrong! */
1142 AUDIT_ARG_PID((pid_t)id); /* XXX - This may be wrong! */
1143 AUDIT_ARG_VALUE(options);
1147 if ((pid_t)id == WAIT_MYPGRP && (idtype == P_PID || idtype == P_PGID)) {
1149 id = (id_t)q->p_pgid;
1154 /* If we don't know the option, just return. */
1155 if ((options & ~(WUNTRACED | WNOHANG | WCONTINUED | WNOWAIT |
1156 WEXITED | WTRAPPED | WLINUXCLONE)) != 0)
1158 if ((options & (WEXITED | WUNTRACED | WCONTINUED | WTRAPPED)) == 0) {
1160 * We will be unable to find any matching processes,
1161 * because there are no known events to look for.
1162 * Prefer to return error instead of blocking
1169 if (q->p_flag & P_STATCHILD) {
1171 q->p_flag &= ~P_STATCHILD;
1175 sx_xlock(&proctree_lock);
1176 LIST_FOREACH(p, &q->p_children, p_sibling) {
1177 ret = proc_to_reap(td, p, idtype, id, status, options,
1189 if ((options & WTRAPPED) != 0 &&
1190 (p->p_flag & P_TRACED) != 0 &&
1191 (p->p_flag & (P_STOPPED_TRACE | P_STOPPED_SIG)) != 0 &&
1192 (p->p_suspcount == p->p_numthreads) &&
1193 ((p->p_flag & P_WAITED) == 0)) {
1195 if ((options & WNOWAIT) == 0)
1196 p->p_flag |= P_WAITED;
1197 sx_xunlock(&proctree_lock);
1198 td->td_retval[0] = p->p_pid;
1201 *status = W_STOPCODE(p->p_xstat);
1202 if (siginfo != NULL) {
1203 siginfo->si_status = p->p_xstat;
1204 siginfo->si_code = CLD_TRAPPED;
1206 if ((options & WNOWAIT) == 0) {
1208 sigqueue_take(p->p_ksi);
1215 if ((options & WUNTRACED) != 0 &&
1216 (p->p_flag & P_STOPPED_SIG) != 0 &&
1217 (p->p_suspcount == p->p_numthreads) &&
1218 ((p->p_flag & P_WAITED) == 0)) {
1220 if ((options & WNOWAIT) == 0)
1221 p->p_flag |= P_WAITED;
1222 sx_xunlock(&proctree_lock);
1223 td->td_retval[0] = p->p_pid;
1226 *status = W_STOPCODE(p->p_xstat);
1227 if (siginfo != NULL) {
1228 siginfo->si_status = p->p_xstat;
1229 siginfo->si_code = CLD_STOPPED;
1231 if ((options & WNOWAIT) == 0) {
1233 sigqueue_take(p->p_ksi);
1241 if ((options & WCONTINUED) != 0 &&
1242 (p->p_flag & P_CONTINUED) != 0) {
1243 sx_xunlock(&proctree_lock);
1244 td->td_retval[0] = p->p_pid;
1245 if ((options & WNOWAIT) == 0) {
1246 p->p_flag &= ~P_CONTINUED;
1248 sigqueue_take(p->p_ksi);
1255 if (siginfo != NULL) {
1256 siginfo->si_status = SIGCONT;
1257 siginfo->si_code = CLD_CONTINUED;
1265 * Look in the orphans list too, to allow the parent to
1266 * collect it's child exit status even if child is being
1269 * Debugger detaches from the parent upon successful
1270 * switch-over from parent to child. At this point due to
1271 * re-parenting the parent loses the child to debugger and a
1272 * wait4(2) call would report that it has no children to wait
1273 * for. By maintaining a list of orphans we allow the parent
1274 * to successfully wait until the child becomes a zombie.
1276 LIST_FOREACH(p, &q->p_orphans, p_orphan) {
1277 ret = proc_to_reap(td, p, idtype, id, status, options,
1287 sx_xunlock(&proctree_lock);
1290 if (options & WNOHANG) {
1291 sx_xunlock(&proctree_lock);
1292 td->td_retval[0] = 0;
1296 sx_xunlock(&proctree_lock);
1297 if (q->p_flag & P_STATCHILD) {
1298 q->p_flag &= ~P_STATCHILD;
1301 error = msleep(q, &q->p_mtx, PWAIT | PCATCH, "wait", 0);
1309 * Make process 'parent' the new parent of process 'child'.
1310 * Must be called with an exclusive hold of proctree lock.
1313 proc_reparent(struct proc *child, struct proc *parent)
1316 sx_assert(&proctree_lock, SX_XLOCKED);
1317 PROC_LOCK_ASSERT(child, MA_OWNED);
1318 if (child->p_pptr == parent)
1321 PROC_LOCK(child->p_pptr);
1322 sigqueue_take(child->p_ksi);
1323 PROC_UNLOCK(child->p_pptr);
1324 LIST_REMOVE(child, p_sibling);
1325 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
1327 clear_orphan(child);
1328 if (child->p_flag & P_TRACED) {
1329 if (LIST_EMPTY(&child->p_pptr->p_orphans)) {
1330 child->p_treeflag |= P_TREE_FIRST_ORPHAN;
1331 LIST_INSERT_HEAD(&child->p_pptr->p_orphans, child,
1334 LIST_INSERT_AFTER(LIST_FIRST(&child->p_pptr->p_orphans),
1337 child->p_treeflag |= P_TREE_ORPHANED;
1340 child->p_pptr = parent;