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/capsicum.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>
78 #include <sys/ktrace.h>
81 #include <security/audit/audit.h>
82 #include <security/mac/mac_framework.h>
85 #include <vm/vm_extern.h>
86 #include <vm/vm_param.h>
88 #include <vm/vm_map.h>
89 #include <vm/vm_page.h>
93 #include <sys/dtrace_bsd.h>
94 dtrace_execexit_func_t dtrace_fasttrap_exit;
97 SDT_PROVIDER_DECLARE(proc);
98 SDT_PROBE_DEFINE1(proc, kernel, , exit, "int");
100 /* Hook for NFS teardown procedure. */
101 void (*nlminfo_release_p)(struct proc *p);
104 proc_realparent(struct proc *child)
106 struct proc *p, *parent;
108 sx_assert(&proctree_lock, SX_LOCKED);
109 if ((child->p_treeflag & P_TREE_ORPHANED) == 0) {
110 if (child->p_oppid == 0 ||
111 child->p_pptr->p_pid == child->p_oppid)
112 parent = child->p_pptr;
117 for (p = child; (p->p_treeflag & P_TREE_FIRST_ORPHAN) == 0;) {
118 /* Cannot use LIST_PREV(), since the list head is not known. */
119 p = __containerof(p->p_orphan.le_prev, struct proc,
121 KASSERT((p->p_treeflag & P_TREE_ORPHANED) != 0,
122 ("missing P_ORPHAN %p", p));
124 parent = __containerof(p->p_orphan.le_prev, struct proc,
130 reaper_abandon_children(struct proc *p, bool exiting)
132 struct proc *p1, *p2, *ptmp;
134 sx_assert(&proctree_lock, SX_LOCKED);
135 KASSERT(p != initproc, ("reaper_abandon_children for initproc"));
136 if ((p->p_treeflag & P_TREE_REAPER) == 0)
139 LIST_FOREACH_SAFE(p2, &p->p_reaplist, p_reapsibling, ptmp) {
140 LIST_REMOVE(p2, p_reapsibling);
142 p2->p_reapsubtree = p->p_reapsubtree;
143 LIST_INSERT_HEAD(&p1->p_reaplist, p2, p_reapsibling);
144 if (exiting && p2->p_pptr == p) {
146 proc_reparent(p2, p1);
150 KASSERT(LIST_EMPTY(&p->p_reaplist), ("p_reaplist not empty"));
151 p->p_treeflag &= ~P_TREE_REAPER;
155 clear_orphan(struct proc *p)
159 sx_assert(&proctree_lock, SA_XLOCKED);
160 if ((p->p_treeflag & P_TREE_ORPHANED) == 0)
162 if ((p->p_treeflag & P_TREE_FIRST_ORPHAN) != 0) {
163 p1 = LIST_NEXT(p, p_orphan);
165 p1->p_treeflag |= P_TREE_FIRST_ORPHAN;
166 p->p_treeflag &= ~P_TREE_FIRST_ORPHAN;
168 LIST_REMOVE(p, p_orphan);
169 p->p_treeflag &= ~P_TREE_ORPHANED;
173 * exit -- death of process.
176 sys_sys_exit(struct thread *td, struct sys_exit_args *uap)
179 exit1(td, W_EXITCODE(uap->rval, 0));
184 * Exit: deallocate address space and other resources, change proc state to
185 * zombie, and unlink proc from allproc and parent's lists. Save exit status
186 * and rusage for wait(). Check for child processes and orphan them.
189 exit1(struct thread *td, int rv)
191 struct proc *p, *nq, *q, *t;
193 struct vnode *ttyvp = NULL;
195 mtx_assert(&Giant, MA_NOTOWNED);
199 * XXX in case we're rebooting we just let init die in order to
200 * work around an unsolved stack overflow seen very late during
201 * shutdown on sparc64 when the gmirror worker process exists.
203 if (p == initproc && rebooting == 0) {
204 printf("init died (signal %d, exit %d)\n",
205 WTERMSIG(rv), WEXITSTATUS(rv));
206 panic("Going nowhere without my init!");
210 * Deref SU mp, since the thread does not return to userspace.
212 if (softdep_ast_cleanup != NULL)
213 softdep_ast_cleanup();
216 * MUST abort all other threads before proceeding past here.
220 * First check if some other thread or external request got
221 * here before us. If so, act appropriately: exit or suspend.
222 * We must ensure that stop requests are handled before we set
225 thread_suspend_check(0);
226 while (p->p_flag & P_HADTHREADS) {
228 * Kill off the other threads. This requires
229 * some co-operation from other parts of the kernel
230 * so it may not be instantaneous. With this state set
231 * any thread entering the kernel from userspace will
232 * thread_exit() in trap(). Any thread attempting to
233 * sleep will return immediately with EINTR or EWOULDBLOCK
234 * which will hopefully force them to back out to userland
235 * freeing resources as they go. Any thread attempting
236 * to return to userland will thread_exit() from userret().
237 * thread_exit() will unsuspend us when the last of the
238 * other threads exits.
239 * If there is already a thread singler after resumption,
240 * calling thread_single will fail; in that case, we just
241 * re-check all suspension request, the thread should
242 * either be suspended there or exit.
244 if (!thread_single(p, SINGLE_EXIT))
246 * All other activity in this process is now
247 * stopped. Threading support has been turned
252 * Recheck for new stop or suspend requests which
253 * might appear while process lock was dropped in
256 thread_suspend_check(0);
258 KASSERT(p->p_numthreads == 1,
259 ("exit1: proc %p exiting with %d threads", p, p->p_numthreads));
260 racct_sub(p, RACCT_NTHR, 1);
262 * Wakeup anyone in procfs' PIOCWAIT. They should have a hold
263 * on our vmspace, so we should block below until they have
264 * released their reference to us. Note that if they have
265 * requested S_EXIT stops we will block here until they ack
268 _STOPEVENT(p, S_EXIT, rv);
271 * Ignore any pending request to stop due to a stop signal.
272 * Once P_WEXIT is set, future requests will be ignored as
275 p->p_flag &= ~P_STOPPED_SIG;
276 KASSERT(!P_SHOULDSTOP(p), ("exiting process is stopped"));
279 * Note that we are exiting and do another wakeup of anyone in
280 * PIOCWAIT in case they aren't listening for S_EXIT stops or
281 * decided to wait again after we told them we are exiting.
283 p->p_flag |= P_WEXIT;
287 * Wait for any processes that have a hold on our vmspace to
288 * release their reference.
290 while (p->p_lock > 0)
291 msleep(&p->p_lock, &p->p_mtx, PWAIT, "exithold", 0);
293 p->p_xstat = rv; /* Let event handler change exit status */
295 /* Drain the limit callout while we don't have the proc locked */
296 callout_drain(&p->p_limco);
300 * The Sun BSM exit token contains two components: an exit status as
301 * passed to exit(), and a return value to indicate what sort of exit
302 * it was. The exit status is WEXITSTATUS(rv), but it's not clear
303 * what the return value is.
305 AUDIT_ARG_EXIT(WEXITSTATUS(rv), 0);
306 AUDIT_SYSCALL_EXIT(0, td);
309 /* Are we a task leader? */
310 if (p == p->p_leader) {
311 mtx_lock(&ppeers_lock);
315 kern_psignal(q, SIGKILL);
319 while (p->p_peers != NULL)
320 msleep(p, &ppeers_lock, PWAIT, "exit1", 0);
321 mtx_unlock(&ppeers_lock);
325 * Check if any loadable modules need anything done at process exit.
326 * E.g. SYSV IPC stuff
327 * XXX what if one of these generates an error?
329 EVENTHANDLER_INVOKE(process_exit, p);
332 * If parent is waiting for us to exit or exec,
333 * P_PPWAIT is set; we will wakeup the parent below.
336 rv = p->p_xstat; /* Event handler could change exit status */
338 p->p_flag &= ~(P_TRACED | P_PPWAIT | P_PPTRACE);
341 * Stop the real interval timer. If the handler is currently
342 * executing, prevent it from rearming itself and let it finish.
344 if (timevalisset(&p->p_realtimer.it_value) &&
345 callout_stop(&p->p_itcallout) == 0) {
346 timevalclear(&p->p_realtimer.it_interval);
347 msleep(&p->p_itcallout, &p->p_mtx, PWAIT, "ritwait", 0);
348 KASSERT(!timevalisset(&p->p_realtimer.it_value),
349 ("realtime timer is still armed"));
354 * Reset any sigio structures pointing to us as a result of
355 * F_SETOWN with our pid.
357 funsetownlst(&p->p_sigiolst);
360 * If this process has an nlminfo data area (for lockd), release it
362 if (nlminfo_release_p != NULL && p->p_nlminfo != NULL)
363 (*nlminfo_release_p)(p);
366 * Close open files and release open-file table.
372 * If this thread tickled GEOM, we need to wait for the giggling to
373 * stop before we return to userland
375 if (td->td_pflags & TDP_GEOM)
379 * Remove ourself from our leader's peer list and wake our leader.
381 mtx_lock(&ppeers_lock);
382 if (p->p_leader->p_peers) {
384 while (q->p_peers != p)
386 q->p_peers = p->p_peers;
389 mtx_unlock(&ppeers_lock);
393 sx_xlock(&proctree_lock);
394 if (SESS_LEADER(p)) {
395 struct session *sp = p->p_session;
399 * s_ttyp is not zero'd; we use this to indicate that
400 * the session once had a controlling terminal. (for
401 * logging and informational purposes)
412 * Signal foreground pgrp and revoke access to
413 * controlling terminal if it has not been revoked
416 * Because the TTY may have been revoked in the mean
417 * time and could already have a new session associated
418 * with it, make sure we don't send a SIGHUP to a
419 * foreground process group that does not belong to this
425 if (tp->t_session == sp)
426 tty_signal_pgrp(tp, SIGHUP);
431 sx_xunlock(&proctree_lock);
432 if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
433 VOP_REVOKE(ttyvp, REVOKEALL);
434 VOP_UNLOCK(ttyvp, 0);
436 sx_xlock(&proctree_lock);
439 fixjobc(p, p->p_pgrp, 0);
440 sx_xunlock(&proctree_lock);
441 (void)acct_process(td);
443 /* Release the TTY now we've unlocked everything. */
450 * Release reference to text vnode
452 if (p->p_textvp != NULL) {
458 * Release our limits structure.
460 lim_free(p->p_limit);
466 * Remove proc from allproc queue and pidhash chain.
467 * Place onto zombproc. Unlink from parent's child list.
469 sx_xlock(&allproc_lock);
470 LIST_REMOVE(p, p_list);
471 LIST_INSERT_HEAD(&zombproc, p, p_list);
472 LIST_REMOVE(p, p_hash);
473 sx_xunlock(&allproc_lock);
476 * Call machine-dependent code to release any
477 * machine-dependent resources other than the address space.
478 * The address space is released by "vmspace_exitfree(p)" in
483 WITNESS_WARN(WARN_PANIC, NULL, "process (pid %d) exiting", p->p_pid);
486 * Reparent all children processes:
487 * - traced ones to the original parent (or init if we are that parent)
490 sx_xlock(&proctree_lock);
491 q = LIST_FIRST(&p->p_children);
492 if (q != NULL) /* only need this if any child is S_ZOMB */
494 for (; q != NULL; q = nq) {
495 nq = LIST_NEXT(q, p_sibling);
497 q->p_sigparent = SIGCHLD;
499 if (!(q->p_flag & P_TRACED)) {
500 proc_reparent(q, q->p_reaper);
503 * Traced processes are killed since their existence
504 * means someone is screwing up.
506 t = proc_realparent(q);
508 proc_reparent(q, q->p_reaper);
515 * Since q was found on our children list, the
516 * proc_reparent() call moved q to the orphan
517 * list due to present P_TRACED flag. Clear
518 * orphan link for q now while q is locked.
521 q->p_flag &= ~(P_TRACED | P_STOPPED_TRACE);
522 FOREACH_THREAD_IN_PROC(q, tdt)
523 tdt->td_dbgflags &= ~TDB_SUSPEND;
524 kern_psignal(q, SIGKILL);
530 * Also get rid of our orphans.
532 while ((q = LIST_FIRST(&p->p_orphans)) != NULL) {
534 CTR2(KTR_PTRACE, "exit: pid %d, clearing orphan %d", p->p_pid,
540 /* Save exit status. */
544 /* Tell the prison that we are gone. */
545 prison_proc_free(p->p_ucred->cr_prison);
549 * Tell the DTrace fasttrap provider about the exit if it
550 * has declared an interest.
552 if (dtrace_fasttrap_exit)
553 dtrace_fasttrap_exit(p);
557 * Notify interested parties of our demise.
559 KNOTE_LOCKED(&p->p_klist, NOTE_EXIT);
562 int reason = CLD_EXITED;
565 else if (WIFSIGNALED(rv))
567 SDT_PROBE(proc, kernel, , exit, reason, 0, 0, 0, 0);
571 * Just delete all entries in the p_klist. At this point we won't
572 * report any more events, and there are nasty race conditions that
573 * can beat us if we don't.
575 knlist_clear(&p->p_klist, 1);
578 * If this is a process with a descriptor, we may not need to deliver
579 * a signal to the parent. proctree_lock is held over
580 * procdesc_exit() to serialize concurrent calls to close() and
584 if (p->p_procdesc == NULL || procdesc_exit(p)) {
587 * Notify parent that we're gone. If parent has the
588 * PS_NOCLDWAIT flag set, or if the handler is set to SIG_IGN,
589 * notify process 1 instead (and hope it will handle this
592 PROC_LOCK(p->p_pptr);
593 mtx_lock(&p->p_pptr->p_sigacts->ps_mtx);
594 if (p->p_pptr->p_sigacts->ps_flag &
595 (PS_NOCLDWAIT | PS_CLDSIGIGN)) {
598 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
601 proc_reparent(p, p->p_reaper);
602 p->p_sigparent = SIGCHLD;
603 PROC_LOCK(p->p_pptr);
606 * Notify parent, so in case he was wait(2)ing or
607 * executing waitpid(2) with our pid, he will
612 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
614 if (p->p_pptr == p->p_reaper || p->p_pptr == initproc)
616 else if (p->p_sigparent != 0) {
617 if (p->p_sigparent == SIGCHLD)
619 else /* LINUX thread */
620 kern_psignal(p->p_pptr, p->p_sigparent);
624 PROC_LOCK(p->p_pptr);
626 sx_xunlock(&proctree_lock);
629 * The state PRS_ZOMBIE prevents other proesses from sending
630 * signal to the process, to avoid memory leak, we free memory
631 * for signal queue at the time when the state is set.
633 sigqueue_flush(&p->p_sigqueue);
634 sigqueue_flush(&td->td_sigqueue);
637 * We have to wait until after acquiring all locks before
638 * changing p_state. We need to avoid all possible context
639 * switches (including ones from blocking on a mutex) while
640 * marked as a zombie. We also have to set the zombie state
641 * before we release the parent process' proc lock to avoid
642 * a lost wakeup. So, we first call wakeup, then we grab the
643 * sched lock, update the state, and release the parent process'
647 cv_broadcast(&p->p_pwait);
648 sched_exit(p->p_pptr, td);
649 umtx_thread_exit(td);
651 p->p_state = PRS_ZOMBIE;
652 PROC_UNLOCK(p->p_pptr);
655 * Hopefully no one will try to deliver a signal to the process this
658 knlist_destroy(&p->p_klist);
661 * Save our children's rusage information in our exit rusage.
663 ruadd(&p->p_ru, &p->p_rux, &p->p_stats->p_cru, &p->p_crux);
666 * Make sure the scheduler takes this thread out of its tables etc.
667 * This will also release this thread's reference to the ucred.
668 * Other thread parts to release include pcb bits and such.
674 #ifndef _SYS_SYSPROTO_H_
683 sys_abort2(struct thread *td, struct abort2_args *uap)
685 struct proc *p = td->td_proc;
691 * Do it right now so we can log either proper call of abort2(), or
692 * note, that invalid argument was passed. 512 is big enough to
693 * handle 16 arguments' descriptions with additional comments.
695 sb = sbuf_new(NULL, NULL, 512, SBUF_FIXEDLEN);
697 sbuf_printf(sb, "%s(pid %d uid %d) aborted: ",
698 p->p_comm, p->p_pid, td->td_ucred->cr_uid);
700 * Since we can't return from abort2(), send SIGKILL in cases, where
701 * abort2() was called improperly
704 /* Prevent from DoSes from user-space. */
705 if (uap->nargs < 0 || uap->nargs > 16)
707 if (uap->nargs > 0) {
708 if (uap->args == NULL)
710 error = copyin(uap->args, uargs, uap->nargs * sizeof(void *));
715 * Limit size of 'reason' string to 128. Will fit even when
716 * maximal number of arguments was chosen to be logged.
718 if (uap->why != NULL) {
719 error = sbuf_copyin(sb, uap->why, 128);
723 sbuf_printf(sb, "(null)");
725 if (uap->nargs > 0) {
726 sbuf_printf(sb, "(");
727 for (i = 0;i < uap->nargs; i++)
728 sbuf_printf(sb, "%s%p", i == 0 ? "" : ", ", uargs[i]);
729 sbuf_printf(sb, ")");
732 * Final stage: arguments were proper, string has been
733 * successfully copied from userspace, and copying pointers
734 * from user-space succeed.
738 if (sig == SIGKILL) {
740 sbuf_printf(sb, " (Reason text inaccessible)");
744 log(LOG_INFO, "%s", sbuf_data(sb));
746 exit1(td, W_EXITCODE(0, sig));
753 * The dirty work is handled by kern_wait().
756 owait(struct thread *td, struct owait_args *uap __unused)
760 error = kern_wait(td, WAIT_ANY, &status, 0, NULL);
762 td->td_retval[1] = status;
765 #endif /* COMPAT_43 */
768 * The dirty work is handled by kern_wait().
771 sys_wait4(struct thread *td, struct wait4_args *uap)
773 struct rusage ru, *rup;
776 if (uap->rusage != NULL)
780 error = kern_wait(td, uap->pid, &status, uap->options, rup);
781 if (uap->status != NULL && error == 0)
782 error = copyout(&status, uap->status, sizeof(status));
783 if (uap->rusage != NULL && error == 0)
784 error = copyout(&ru, uap->rusage, sizeof(struct rusage));
789 sys_wait6(struct thread *td, struct wait6_args *uap)
791 struct __wrusage wru, *wrup;
797 idtype = uap->idtype;
800 if (uap->wrusage != NULL)
805 if (uap->info != NULL) {
807 bzero(sip, sizeof(*sip));
812 * We expect all callers of wait6() to know about WEXITED and
815 error = kern_wait6(td, idtype, id, &status, uap->options, wrup, sip);
817 if (uap->status != NULL && error == 0)
818 error = copyout(&status, uap->status, sizeof(status));
819 if (uap->wrusage != NULL && error == 0)
820 error = copyout(&wru, uap->wrusage, sizeof(wru));
821 if (uap->info != NULL && error == 0)
822 error = copyout(&si, uap->info, sizeof(si));
827 * Reap the remains of a zombie process and optionally return status and
828 * rusage. Asserts and will release both the proctree_lock and the process
829 * lock as part of its work.
832 proc_reap(struct thread *td, struct proc *p, int *status, int options)
836 sx_assert(&proctree_lock, SA_XLOCKED);
837 PROC_LOCK_ASSERT(p, MA_OWNED);
838 PROC_SLOCK_ASSERT(p, MA_OWNED);
839 KASSERT(p->p_state == PRS_ZOMBIE, ("proc_reap: !PRS_ZOMBIE"));
844 td->td_retval[0] = p->p_pid;
846 *status = p->p_xstat; /* convert to int */
847 if (options & WNOWAIT) {
849 * Only poll, returning the status. Caller does not wish to
850 * release the proc struct just yet.
853 sx_xunlock(&proctree_lock);
858 sigqueue_take(p->p_ksi);
863 * If we got the child via a ptrace 'attach', we need to give it back
866 if (p->p_oppid != 0) {
867 t = proc_realparent(p);
871 "wait: traced child %d moved back to parent %d", p->p_pid,
876 pksignal(t, SIGCHLD, p->p_ksi);
878 cv_broadcast(&p->p_pwait);
880 sx_xunlock(&proctree_lock);
885 * Remove other references to this process to ensure we have an
886 * exclusive reference.
888 sx_xlock(&allproc_lock);
889 LIST_REMOVE(p, p_list); /* off zombproc */
890 sx_xunlock(&allproc_lock);
891 LIST_REMOVE(p, p_sibling);
892 reaper_abandon_children(p, true);
893 LIST_REMOVE(p, p_reapsibling);
899 if (p->p_procdesc != NULL)
902 sx_xunlock(&proctree_lock);
905 * As a side effect of this lock, we know that all other writes to
906 * this proc are visible now, so no more locking is needed for p.
909 p->p_xstat = 0; /* XXX: why? */
912 ruadd(&q->p_stats->p_cru, &q->p_crux, &p->p_ru, &p->p_rux);
916 * Decrement the count of procs running with this uid.
918 (void)chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0);
921 * Destroy resource accounting information associated with the process.
926 racct_sub(p, RACCT_NPROC, 1);
933 * Free credentials, arguments, and sigacts.
937 pargs_drop(p->p_args);
939 sigacts_free(p->p_sigacts);
943 * Do any thread-system specific cleanups.
948 * Give vm and machine-dependent layer a chance to free anything that
949 * cpu_exit couldn't release while still running in process context.
955 KASSERT(FIRST_THREAD_IN_PROC(p),
956 ("proc_reap: no residual thread!"));
957 uma_zfree(proc_zone, p);
958 sx_xlock(&allproc_lock);
960 sx_xunlock(&allproc_lock);
964 proc_to_reap(struct thread *td, struct proc *p, idtype_t idtype, id_t id,
965 int *status, int options, struct __wrusage *wrusage, siginfo_t *siginfo)
969 sx_assert(&proctree_lock, SA_XLOCKED);
977 if (p->p_pid != (pid_t)id) {
983 if (p->p_pgid != (pid_t)id) {
989 if (p->p_session->s_sid != (pid_t)id) {
995 if (p->p_ucred->cr_uid != (uid_t)id) {
1001 if (p->p_ucred->cr_gid != (gid_t)id) {
1007 if (p->p_ucred->cr_prison->pr_id != (int)id) {
1013 * It seems that the thread structures get zeroed out
1014 * at process exit. This makes it impossible to
1015 * support P_SETID, P_CID or P_CPUID.
1022 if (p_canwait(td, p)) {
1027 if (((options & WEXITED) == 0) && (p->p_state == PRS_ZOMBIE)) {
1033 * This special case handles a kthread spawned by linux_clone
1034 * (see linux_misc.c). The linux_wait4 and linux_waitpid
1035 * functions need to be able to distinguish between waiting
1036 * on a process and waiting on a thread. It is a thread if
1037 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option
1038 * signifies we want to wait for threads and not processes.
1040 if ((p->p_sigparent != SIGCHLD) ^
1041 ((options & WLINUXCLONE) != 0)) {
1048 if (siginfo != NULL) {
1049 bzero(siginfo, sizeof(*siginfo));
1050 siginfo->si_errno = 0;
1053 * SUSv4 requires that the si_signo value is always
1054 * SIGCHLD. Obey it despite the rfork(2) interface
1055 * allows to request other signal for child exit
1058 siginfo->si_signo = SIGCHLD;
1061 * This is still a rough estimate. We will fix the
1062 * cases TRAPPED, STOPPED, and CONTINUED later.
1064 if (WCOREDUMP(p->p_xstat)) {
1065 siginfo->si_code = CLD_DUMPED;
1066 siginfo->si_status = WTERMSIG(p->p_xstat);
1067 } else if (WIFSIGNALED(p->p_xstat)) {
1068 siginfo->si_code = CLD_KILLED;
1069 siginfo->si_status = WTERMSIG(p->p_xstat);
1071 siginfo->si_code = CLD_EXITED;
1072 siginfo->si_status = WEXITSTATUS(p->p_xstat);
1075 siginfo->si_pid = p->p_pid;
1076 siginfo->si_uid = p->p_ucred->cr_uid;
1079 * The si_addr field would be useful additional
1080 * detail, but apparently the PC value may be lost
1081 * when we reach this point. bzero() above sets
1082 * siginfo->si_addr to NULL.
1087 * There should be no reason to limit resources usage info to
1088 * exited processes only. A snapshot about any resources used
1089 * by a stopped process may be exactly what is needed.
1091 if (wrusage != NULL) {
1092 rup = &wrusage->wru_self;
1094 calcru(p, &rup->ru_utime, &rup->ru_stime);
1096 rup = &wrusage->wru_children;
1097 *rup = p->p_stats->p_cru;
1098 calccru(p, &rup->ru_utime, &rup->ru_stime);
1101 if (p->p_state == PRS_ZOMBIE) {
1102 proc_reap(td, p, status, options);
1111 kern_wait(struct thread *td, pid_t pid, int *status, int options,
1112 struct rusage *rusage)
1114 struct __wrusage wru, *wrup;
1120 * Translate the special pid values into the (idtype, pid)
1121 * pair for kern_wait6. The WAIT_MYPGRP case is handled by
1122 * kern_wait6() on its own.
1124 if (pid == WAIT_ANY) {
1127 } else if (pid < 0) {
1141 * For backward compatibility we implicitly add flags WEXITED
1142 * and WTRAPPED here.
1144 options |= WEXITED | WTRAPPED;
1145 ret = kern_wait6(td, idtype, id, status, options, wrup, NULL);
1147 *rusage = wru.wru_self;
1152 kern_wait6(struct thread *td, idtype_t idtype, id_t id, int *status,
1153 int options, struct __wrusage *wrusage, siginfo_t *siginfo)
1156 int error, nfound, ret;
1158 AUDIT_ARG_VALUE((int)idtype); /* XXX - This is likely wrong! */
1159 AUDIT_ARG_PID((pid_t)id); /* XXX - This may be wrong! */
1160 AUDIT_ARG_VALUE(options);
1164 if ((pid_t)id == WAIT_MYPGRP && (idtype == P_PID || idtype == P_PGID)) {
1166 id = (id_t)q->p_pgid;
1171 /* If we don't know the option, just return. */
1172 if ((options & ~(WUNTRACED | WNOHANG | WCONTINUED | WNOWAIT |
1173 WEXITED | WTRAPPED | WLINUXCLONE)) != 0)
1175 if ((options & (WEXITED | WUNTRACED | WCONTINUED | WTRAPPED)) == 0) {
1177 * We will be unable to find any matching processes,
1178 * because there are no known events to look for.
1179 * Prefer to return error instead of blocking
1186 if (q->p_flag & P_STATCHILD) {
1188 q->p_flag &= ~P_STATCHILD;
1192 sx_xlock(&proctree_lock);
1193 LIST_FOREACH(p, &q->p_children, p_sibling) {
1194 ret = proc_to_reap(td, p, idtype, id, status, options,
1206 if ((options & WTRAPPED) != 0 &&
1207 (p->p_flag & P_TRACED) != 0 &&
1208 (p->p_flag & (P_STOPPED_TRACE | P_STOPPED_SIG)) != 0 &&
1209 (p->p_suspcount == p->p_numthreads) &&
1210 ((p->p_flag & P_WAITED) == 0)) {
1212 if ((options & WNOWAIT) == 0)
1213 p->p_flag |= P_WAITED;
1214 sx_xunlock(&proctree_lock);
1215 td->td_retval[0] = p->p_pid;
1218 *status = W_STOPCODE(p->p_xstat);
1219 if (siginfo != NULL) {
1220 siginfo->si_status = p->p_xstat;
1221 siginfo->si_code = CLD_TRAPPED;
1223 if ((options & WNOWAIT) == 0) {
1225 sigqueue_take(p->p_ksi);
1230 "wait: returning trapped pid %d status %#x (xstat %d) xthread %d",
1231 p->p_pid, W_STOPCODE(p->p_xstat), p->p_xstat,
1232 p->p_xthread != NULL ? p->p_xthread->td_tid : -1);
1236 if ((options & WUNTRACED) != 0 &&
1237 (p->p_flag & P_STOPPED_SIG) != 0 &&
1238 (p->p_suspcount == p->p_numthreads) &&
1239 ((p->p_flag & P_WAITED) == 0)) {
1241 if ((options & WNOWAIT) == 0)
1242 p->p_flag |= P_WAITED;
1243 sx_xunlock(&proctree_lock);
1244 td->td_retval[0] = p->p_pid;
1247 *status = W_STOPCODE(p->p_xstat);
1248 if (siginfo != NULL) {
1249 siginfo->si_status = p->p_xstat;
1250 siginfo->si_code = CLD_STOPPED;
1252 if ((options & WNOWAIT) == 0) {
1254 sigqueue_take(p->p_ksi);
1262 if ((options & WCONTINUED) != 0 &&
1263 (p->p_flag & P_CONTINUED) != 0) {
1264 sx_xunlock(&proctree_lock);
1265 td->td_retval[0] = p->p_pid;
1266 if ((options & WNOWAIT) == 0) {
1267 p->p_flag &= ~P_CONTINUED;
1269 sigqueue_take(p->p_ksi);
1276 if (siginfo != NULL) {
1277 siginfo->si_status = SIGCONT;
1278 siginfo->si_code = CLD_CONTINUED;
1286 * Look in the orphans list too, to allow the parent to
1287 * collect it's child exit status even if child is being
1290 * Debugger detaches from the parent upon successful
1291 * switch-over from parent to child. At this point due to
1292 * re-parenting the parent loses the child to debugger and a
1293 * wait4(2) call would report that it has no children to wait
1294 * for. By maintaining a list of orphans we allow the parent
1295 * to successfully wait until the child becomes a zombie.
1297 LIST_FOREACH(p, &q->p_orphans, p_orphan) {
1298 ret = proc_to_reap(td, p, idtype, id, status, options,
1308 sx_xunlock(&proctree_lock);
1311 if (options & WNOHANG) {
1312 sx_xunlock(&proctree_lock);
1313 td->td_retval[0] = 0;
1317 sx_xunlock(&proctree_lock);
1318 if (q->p_flag & P_STATCHILD) {
1319 q->p_flag &= ~P_STATCHILD;
1322 error = msleep(q, &q->p_mtx, PWAIT | PCATCH, "wait", 0);
1330 * Make process 'parent' the new parent of process 'child'.
1331 * Must be called with an exclusive hold of proctree lock.
1334 proc_reparent(struct proc *child, struct proc *parent)
1337 sx_assert(&proctree_lock, SX_XLOCKED);
1338 PROC_LOCK_ASSERT(child, MA_OWNED);
1339 if (child->p_pptr == parent)
1342 PROC_LOCK(child->p_pptr);
1343 sigqueue_take(child->p_ksi);
1344 PROC_UNLOCK(child->p_pptr);
1345 LIST_REMOVE(child, p_sibling);
1346 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
1348 clear_orphan(child);
1349 if (child->p_flag & P_TRACED) {
1350 if (LIST_EMPTY(&child->p_pptr->p_orphans)) {
1351 child->p_treeflag |= P_TREE_FIRST_ORPHAN;
1352 LIST_INSERT_HEAD(&child->p_pptr->p_orphans, child,
1355 LIST_INSERT_AFTER(LIST_FIRST(&child->p_pptr->p_orphans),
1358 child->p_treeflag |= P_TREE_ORPHANED;
1361 child->p_pptr = parent;