2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1982, 1986, 1989, 1991, 1993
5 * The Regents of the University of California. All rights reserved.
6 * (c) UNIX System Laboratories, Inc.
7 * All or some portions of this file are derived from material licensed
8 * to the University of California by American Telephone and Telegraph
9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10 * the permission of UNIX System Laboratories, Inc.
12 * Redistribution and use in source and binary forms, with or without
13 * modification, are permitted provided that the following conditions
15 * 1. Redistributions of source code must retain the above copyright
16 * notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 * notice, this list of conditions and the following disclaimer in the
19 * documentation and/or other materials provided with the distribution.
20 * 3. Neither the name of the University nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 * @(#)kern_exit.c 8.7 (Berkeley) 2/12/94
39 #include <sys/cdefs.h>
40 __FBSDID("$FreeBSD$");
42 #include "opt_ktrace.h"
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/sysproto.h>
47 #include <sys/capsicum.h>
48 #include <sys/eventhandler.h>
49 #include <sys/kernel.h>
51 #include <sys/malloc.h>
53 #include <sys/mutex.h>
55 #include <sys/procdesc.h>
59 #include <sys/vmmeter.h>
60 #include <sys/vnode.h>
61 #include <sys/racct.h>
62 #include <sys/resourcevar.h>
64 #include <sys/signalvar.h>
65 #include <sys/sched.h>
67 #include <sys/syscallsubr.h>
68 #include <sys/syslog.h>
69 #include <sys/ptrace.h>
70 #include <sys/acct.h> /* for acct_process() function prototype */
71 #include <sys/filedesc.h>
75 #include <sys/timers.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, , , exit, "int");
101 proc_realparent(struct proc *child)
103 struct proc *p, *parent;
105 sx_assert(&proctree_lock, SX_LOCKED);
106 if ((child->p_treeflag & P_TREE_ORPHANED) == 0)
107 return (child->p_pptr->p_pid == child->p_oppid ?
108 child->p_pptr : child->p_reaper);
109 for (p = child; (p->p_treeflag & P_TREE_FIRST_ORPHAN) == 0;) {
110 /* Cannot use LIST_PREV(), since the list head is not known. */
111 p = __containerof(p->p_orphan.le_prev, struct proc,
113 KASSERT((p->p_treeflag & P_TREE_ORPHANED) != 0,
114 ("missing P_ORPHAN %p", p));
116 parent = __containerof(p->p_orphan.le_prev, struct proc,
122 reaper_abandon_children(struct proc *p, bool exiting)
124 struct proc *p1, *p2, *ptmp;
126 sx_assert(&proctree_lock, SX_LOCKED);
127 KASSERT(p != initproc, ("reaper_abandon_children for initproc"));
128 if ((p->p_treeflag & P_TREE_REAPER) == 0)
131 LIST_FOREACH_SAFE(p2, &p->p_reaplist, p_reapsibling, ptmp) {
132 LIST_REMOVE(p2, p_reapsibling);
134 p2->p_reapsubtree = p->p_reapsubtree;
135 LIST_INSERT_HEAD(&p1->p_reaplist, p2, p_reapsibling);
136 if (exiting && p2->p_pptr == p) {
138 proc_reparent(p2, p1, true);
142 KASSERT(LIST_EMPTY(&p->p_reaplist), ("p_reaplist not empty"));
143 p->p_treeflag &= ~P_TREE_REAPER;
147 reaper_clear(struct proc *p)
152 sx_assert(&proctree_lock, SX_LOCKED);
153 LIST_REMOVE(p, p_reapsibling);
154 if (p->p_reapsubtree == 1)
157 LIST_FOREACH(p1, &p->p_reaper->p_reaplist, p_reapsibling) {
158 if (p1->p_reapsubtree == p->p_reapsubtree) {
164 proc_id_clear(PROC_ID_REAP, p->p_reapsubtree);
168 proc_clear_orphan(struct proc *p)
172 sx_assert(&proctree_lock, SA_XLOCKED);
173 if ((p->p_treeflag & P_TREE_ORPHANED) == 0)
175 if ((p->p_treeflag & P_TREE_FIRST_ORPHAN) != 0) {
176 p1 = LIST_NEXT(p, p_orphan);
178 p1->p_treeflag |= P_TREE_FIRST_ORPHAN;
179 p->p_treeflag &= ~P_TREE_FIRST_ORPHAN;
181 LIST_REMOVE(p, p_orphan);
182 p->p_treeflag &= ~P_TREE_ORPHANED;
186 * exit -- death of process.
189 sys_sys_exit(struct thread *td, struct sys_exit_args *uap)
192 exit1(td, uap->rval, 0);
197 * Exit: deallocate address space and other resources, change proc state to
198 * zombie, and unlink proc from allproc and parent's lists. Save exit status
199 * and rusage for wait(). Check for child processes and orphan them.
202 exit1(struct thread *td, int rval, int signo)
204 struct proc *p, *nq, *q, *t;
206 ksiginfo_t *ksi, *ksi1;
209 mtx_assert(&Giant, MA_NOTOWNED);
210 KASSERT(rval == 0 || signo == 0, ("exit1 rv %d sig %d", rval, signo));
214 * XXX in case we're rebooting we just let init die in order to
215 * work around an unsolved stack overflow seen very late during
216 * shutdown on sparc64 when the gmirror worker process exists.
217 * XXX what to do now that sparc64 is gone... remove if?
219 if (p == initproc && rebooting == 0) {
220 printf("init died (signal %d, exit %d)\n", signo, rval);
221 panic("Going nowhere without my init!");
225 * Deref SU mp, since the thread does not return to userspace.
227 td_softdep_cleanup(td);
230 * MUST abort all other threads before proceeding past here.
234 * First check if some other thread or external request got
235 * here before us. If so, act appropriately: exit or suspend.
236 * We must ensure that stop requests are handled before we set
239 thread_suspend_check(0);
240 while (p->p_flag & P_HADTHREADS) {
242 * Kill off the other threads. This requires
243 * some co-operation from other parts of the kernel
244 * so it may not be instantaneous. With this state set
245 * any thread entering the kernel from userspace will
246 * thread_exit() in trap(). Any thread attempting to
247 * sleep will return immediately with EINTR or EWOULDBLOCK
248 * which will hopefully force them to back out to userland
249 * freeing resources as they go. Any thread attempting
250 * to return to userland will thread_exit() from userret().
251 * thread_exit() will unsuspend us when the last of the
252 * other threads exits.
253 * If there is already a thread singler after resumption,
254 * calling thread_single will fail; in that case, we just
255 * re-check all suspension request, the thread should
256 * either be suspended there or exit.
258 if (!thread_single(p, SINGLE_EXIT))
260 * All other activity in this process is now
261 * stopped. Threading support has been turned
266 * Recheck for new stop or suspend requests which
267 * might appear while process lock was dropped in
270 thread_suspend_check(0);
272 KASSERT(p->p_numthreads == 1,
273 ("exit1: proc %p exiting with %d threads", p, p->p_numthreads));
274 racct_sub(p, RACCT_NTHR, 1);
276 /* Let event handler change exit status */
281 * Ignore any pending request to stop due to a stop signal.
282 * Once P_WEXIT is set, future requests will be ignored as
285 p->p_flag &= ~P_STOPPED_SIG;
286 KASSERT(!P_SHOULDSTOP(p), ("exiting process is stopped"));
288 /* Note that we are exiting. */
289 p->p_flag |= P_WEXIT;
292 * Wait for any processes that have a hold on our vmspace to
293 * release their reference.
295 while (p->p_lock > 0)
296 msleep(&p->p_lock, &p->p_mtx, PWAIT, "exithold", 0);
299 /* Drain the limit callout while we don't have the proc locked */
300 callout_drain(&p->p_limco);
304 * The Sun BSM exit token contains two components: an exit status as
305 * passed to exit(), and a return value to indicate what sort of exit
306 * it was. The exit status is WEXITSTATUS(rv), but it's not clear
307 * what the return value is.
309 AUDIT_ARG_EXIT(rval, 0);
310 AUDIT_SYSCALL_EXIT(0, td);
313 /* Are we a task leader with peers? */
314 if (p->p_peers != NULL && p == p->p_leader) {
315 mtx_lock(&ppeers_lock);
319 kern_psignal(q, SIGKILL);
323 while (p->p_peers != NULL)
324 msleep(p, &ppeers_lock, PWAIT, "exit1", 0);
325 mtx_unlock(&ppeers_lock);
331 * Check if any loadable modules need anything done at process exit.
332 * E.g. SYSV IPC stuff.
333 * Event handler could change exit status.
334 * XXX what if one of these generates an error?
336 EVENTHANDLER_DIRECT_INVOKE(process_exit, p);
339 * If parent is waiting for us to exit or exec,
340 * P_PPWAIT is set; we will wakeup the parent below.
347 * Stop the real interval timer. If the handler is currently
348 * executing, prevent it from rearming itself and let it finish.
350 if (timevalisset(&p->p_realtimer.it_value) &&
351 _callout_stop_safe(&p->p_itcallout, CS_EXECUTING, NULL) == 0) {
352 timevalclear(&p->p_realtimer.it_interval);
353 msleep(&p->p_itcallout, &p->p_mtx, PWAIT, "ritwait", 0);
354 KASSERT(!timevalisset(&p->p_realtimer.it_value),
355 ("realtime timer is still armed"));
360 umtx_thread_exit(td);
364 * Reset any sigio structures pointing to us as a result of
365 * F_SETOWN with our pid. The P_WEXIT flag interlocks with fsetown().
367 funsetownlst(&p->p_sigiolst);
370 * Close open files and release open-file table.
377 * If this thread tickled GEOM, we need to wait for the giggling to
378 * stop before we return to userland
380 if (td->td_pflags & TDP_GEOM)
384 * Remove ourself from our leader's peer list and wake our leader.
386 if (p->p_leader->p_peers != NULL) {
387 mtx_lock(&ppeers_lock);
388 if (p->p_leader->p_peers != NULL) {
390 while (q->p_peers != p)
392 q->p_peers = p->p_peers;
395 mtx_unlock(&ppeers_lock);
399 (void)acct_process(td);
405 * Release reference to text vnode
407 if (p->p_textvp != NULL) {
413 * Release our limits structure.
415 lim_free(p->p_limit);
421 * Call machine-dependent code to release any
422 * machine-dependent resources other than the address space.
423 * The address space is released by "vmspace_exitfree(p)" in
428 WITNESS_WARN(WARN_PANIC, NULL, "process (pid %d) exiting", p->p_pid);
431 * Remove from allproc. It still sits in the hash.
433 sx_xlock(&allproc_lock);
434 LIST_REMOVE(p, p_list);
435 sx_xunlock(&allproc_lock);
437 sx_xlock(&proctree_lock);
439 p->p_flag &= ~(P_TRACED | P_PPWAIT | P_PPTRACE);
443 * killjobc() might drop and re-acquire proctree_lock to
444 * revoke control tty if exiting process was a session leader.
449 * Reparent all children processes:
450 * - traced ones to the original parent (or init if we are that parent)
453 q = LIST_FIRST(&p->p_children);
454 if (q != NULL) /* only need this if any child is S_ZOMB */
456 for (; q != NULL; q = nq) {
457 nq = LIST_NEXT(q, p_sibling);
458 ksi = ksiginfo_alloc(TRUE);
460 q->p_sigparent = SIGCHLD;
462 if ((q->p_flag & P_TRACED) == 0) {
463 proc_reparent(q, q->p_reaper, true);
464 if (q->p_state == PRS_ZOMBIE) {
466 * Inform reaper about the reparented
467 * zombie, since wait(2) has something
468 * new to report. Guarantee queueing
469 * of the SIGCHLD signal, similar to
470 * the _exit() behaviour, by providing
471 * our ksiginfo. Ksi is freed by the
474 if (q->p_ksi == NULL) {
477 ksiginfo_copy(q->p_ksi, ksi);
478 ksi->ksi_flags |= KSI_INS;
482 PROC_LOCK(q->p_reaper);
483 pksignal(q->p_reaper, SIGCHLD, ksi1);
484 PROC_UNLOCK(q->p_reaper);
485 } else if (q->p_pdeathsig > 0) {
487 * The child asked to received a signal
490 kern_psignal(q, q->p_pdeathsig);
494 * Traced processes are killed since their existence
495 * means someone is screwing up.
497 t = proc_realparent(q);
499 proc_reparent(q, q->p_reaper, true);
502 proc_reparent(q, t, true);
506 * Since q was found on our children list, the
507 * proc_reparent() call moved q to the orphan
508 * list due to present P_TRACED flag. Clear
509 * orphan link for q now while q is locked.
511 proc_clear_orphan(q);
512 q->p_flag &= ~(P_TRACED | P_STOPPED_TRACE);
513 q->p_flag2 &= ~P2_PTRACE_FSTP;
515 FOREACH_THREAD_IN_PROC(q, tdt) {
516 tdt->td_dbgflags &= ~(TDB_SUSPEND | TDB_XSIG |
519 kern_psignal(q, SIGKILL);
527 * Also get rid of our orphans.
529 while ((q = LIST_FIRST(&p->p_orphans)) != NULL) {
531 KASSERT(q->p_oppid == p->p_pid,
532 ("orphan %p of %p has unexpected oppid %d", q, p,
534 q->p_oppid = q->p_reaper->p_pid;
537 * If we are the real parent of this process
538 * but it has been reparented to a debugger, then
539 * check if it asked for a signal when we exit.
541 if (q->p_pdeathsig > 0)
542 kern_psignal(q, q->p_pdeathsig);
543 CTR2(KTR_PTRACE, "exit: pid %d, clearing orphan %d", p->p_pid,
545 proc_clear_orphan(q);
550 if (SDT_PROBES_ENABLED()) {
551 int reason = CLD_EXITED;
552 if (WCOREDUMP(signo))
554 else if (WIFSIGNALED(signo))
556 SDT_PROBE1(proc, , , exit, reason);
560 /* Save exit status. */
566 * Tell the DTrace fasttrap provider about the exit if it
567 * has declared an interest.
569 if (dtrace_fasttrap_exit)
570 dtrace_fasttrap_exit(p);
574 * Notify interested parties of our demise.
576 KNOTE_LOCKED(p->p_klist, NOTE_EXIT);
579 * If this is a process with a descriptor, we may not need to deliver
580 * a signal to the parent. proctree_lock is held over
581 * procdesc_exit() to serialize concurrent calls to close() and
585 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, true);
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 */
624 PROC_LOCK(p->p_pptr);
625 sx_xunlock(&proctree_lock);
627 if (signal_parent == 1) {
629 } else if (signal_parent == 2) {
630 kern_psignal(p->p_pptr, p->p_sigparent);
633 /* Tell the prison that we are gone. */
634 prison_proc_free(p->p_ucred->cr_prison);
637 * The state PRS_ZOMBIE prevents other proesses from sending
638 * signal to the process, to avoid memory leak, we free memory
639 * for signal queue at the time when the state is set.
641 sigqueue_flush(&p->p_sigqueue);
642 sigqueue_flush(&td->td_sigqueue);
645 * We have to wait until after acquiring all locks before
646 * changing p_state. We need to avoid all possible context
647 * switches (including ones from blocking on a mutex) while
648 * marked as a zombie. We also have to set the zombie state
649 * before we release the parent process' proc lock to avoid
650 * a lost wakeup. So, we first call wakeup, then we grab the
651 * sched lock, update the state, and release the parent process'
655 cv_broadcast(&p->p_pwait);
656 sched_exit(p->p_pptr, td);
658 p->p_state = PRS_ZOMBIE;
659 PROC_UNLOCK(p->p_pptr);
662 * Save our children's rusage information in our exit rusage.
665 ruadd(&p->p_ru, &p->p_rux, &p->p_stats->p_cru, &p->p_crux);
669 * Make sure the scheduler takes this thread out of its tables etc.
670 * This will also release this thread's reference to the ucred.
671 * Other thread parts to release include pcb bits and such.
676 #ifndef _SYS_SYSPROTO_H_
685 sys_abort2(struct thread *td, struct abort2_args *uap)
687 struct proc *p = td->td_proc;
693 * Do it right now so we can log either proper call of abort2(), or
694 * note, that invalid argument was passed. 512 is big enough to
695 * handle 16 arguments' descriptions with additional comments.
697 sb = sbuf_new(NULL, NULL, 512, SBUF_FIXEDLEN);
699 sbuf_printf(sb, "%s(pid %d uid %d) aborted: ",
700 p->p_comm, p->p_pid, td->td_ucred->cr_uid);
702 * Since we can't return from abort2(), send SIGKILL in cases, where
703 * abort2() was called improperly
706 /* Prevent from DoSes from user-space. */
707 if (uap->nargs < 0 || uap->nargs > 16)
709 if (uap->nargs > 0) {
710 if (uap->args == NULL)
712 error = copyin(uap->args, uargs, uap->nargs * sizeof(void *));
717 * Limit size of 'reason' string to 128. Will fit even when
718 * maximal number of arguments was chosen to be logged.
720 if (uap->why != NULL) {
721 error = sbuf_copyin(sb, uap->why, 128);
725 sbuf_printf(sb, "(null)");
727 if (uap->nargs > 0) {
728 sbuf_printf(sb, "(");
729 for (i = 0;i < uap->nargs; i++)
730 sbuf_printf(sb, "%s%p", i == 0 ? "" : ", ", uargs[i]);
731 sbuf_printf(sb, ")");
734 * Final stage: arguments were proper, string has been
735 * successfully copied from userspace, and copying pointers
736 * from user-space succeed.
740 if (sig == SIGKILL) {
742 sbuf_printf(sb, " (Reason text inaccessible)");
746 log(LOG_INFO, "%s", sbuf_data(sb));
754 * The dirty work is handled by kern_wait().
757 owait(struct thread *td, struct owait_args *uap __unused)
761 error = kern_wait(td, WAIT_ANY, &status, 0, NULL);
763 td->td_retval[1] = status;
766 #endif /* COMPAT_43 */
769 * The dirty work is handled by kern_wait().
772 sys_wait4(struct thread *td, struct wait4_args *uap)
774 struct rusage ru, *rup;
777 if (uap->rusage != NULL)
781 error = kern_wait(td, uap->pid, &status, uap->options, rup);
782 if (uap->status != NULL && error == 0 && td->td_retval[0] != 0)
783 error = copyout(&status, uap->status, sizeof(status));
784 if (uap->rusage != NULL && error == 0 && td->td_retval[0] != 0)
785 error = copyout(&ru, uap->rusage, sizeof(struct rusage));
790 sys_wait6(struct thread *td, struct wait6_args *uap)
792 struct __wrusage wru, *wrup;
798 idtype = uap->idtype;
801 if (uap->wrusage != NULL)
806 if (uap->info != NULL) {
808 bzero(sip, sizeof(*sip));
813 * We expect all callers of wait6() to know about WEXITED and
816 error = kern_wait6(td, idtype, id, &status, uap->options, wrup, sip);
818 if (uap->status != NULL && error == 0 && td->td_retval[0] != 0)
819 error = copyout(&status, uap->status, sizeof(status));
820 if (uap->wrusage != NULL && error == 0 && td->td_retval[0] != 0)
821 error = copyout(&wru, uap->wrusage, sizeof(wru));
822 if (uap->info != NULL && error == 0)
823 error = copyout(&si, uap->info, sizeof(si));
828 * Reap the remains of a zombie process and optionally return status and
829 * rusage. Asserts and will release both the proctree_lock and the process
830 * lock as part of its work.
833 proc_reap(struct thread *td, struct proc *p, int *status, int options)
837 sx_assert(&proctree_lock, SA_XLOCKED);
838 PROC_LOCK_ASSERT(p, MA_OWNED);
839 KASSERT(p->p_state == PRS_ZOMBIE, ("proc_reap: !PRS_ZOMBIE"));
841 mtx_spin_wait_unlocked(&p->p_slock);
846 *status = KW_EXITCODE(p->p_xexit, p->p_xsig);
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);
862 * If we got the child via a ptrace 'attach', we need to give it back
865 if (p->p_oppid != p->p_pptr->p_pid) {
867 t = proc_realparent(p);
871 "wait: traced child %d moved back to parent %d", p->p_pid,
873 proc_reparent(p, t, false);
875 pksignal(t, SIGCHLD, p->p_ksi);
877 cv_broadcast(&p->p_pwait);
879 sx_xunlock(&proctree_lock);
885 * Remove other references to this process to ensure we have an
886 * exclusive reference.
888 sx_xlock(PIDHASHLOCK(p->p_pid));
889 LIST_REMOVE(p, p_hash);
890 sx_xunlock(PIDHASHLOCK(p->p_pid));
891 LIST_REMOVE(p, p_sibling);
892 reaper_abandon_children(p, true);
895 proc_clear_orphan(p);
898 if (p->p_procdesc != NULL)
900 sx_xunlock(&proctree_lock);
902 proc_id_clear(PROC_ID_PID, p->p_pid);
905 knlist_detach(p->p_klist);
910 * Removal from allproc list and process group list paired with
911 * PROC_LOCK which was executed during that time should guarantee
912 * nothing can reach this process anymore. As such further locking
915 p->p_xexit = p->p_xsig = 0; /* XXX: why? */
918 ruadd(&q->p_stats->p_cru, &q->p_crux, &p->p_ru, &p->p_rux);
922 * Decrement the count of procs running with this uid.
924 (void)chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0);
927 * Destroy resource accounting information associated with the process.
932 racct_sub(p, RACCT_NPROC, 1);
939 * Free credentials, arguments, and sigacts.
942 pargs_drop(p->p_args);
944 sigacts_free(p->p_sigacts);
948 * Do any thread-system specific cleanups.
953 * Give vm and machine-dependent layer a chance to free anything that
954 * cpu_exit couldn't release while still running in process context.
961 KASSERT(FIRST_THREAD_IN_PROC(p),
962 ("proc_reap: no residual thread!"));
963 uma_zfree(proc_zone, p);
964 atomic_add_int(&nprocs, -1);
968 proc_to_reap(struct thread *td, struct proc *p, idtype_t idtype, id_t id,
969 int *status, int options, struct __wrusage *wrusage, siginfo_t *siginfo,
974 sx_assert(&proctree_lock, SA_XLOCKED);
980 if (p->p_procdesc == NULL ||
981 (p->p_pptr == td->td_proc &&
982 (p->p_flag & P_TRACED) != 0)) {
989 if (p->p_pid != (pid_t)id) {
995 if (p->p_pgid != (pid_t)id) {
1001 if (p->p_session->s_sid != (pid_t)id) {
1007 if (p->p_ucred->cr_uid != (uid_t)id) {
1013 if (p->p_ucred->cr_gid != (gid_t)id) {
1019 if (p->p_ucred->cr_prison->pr_id != (int)id) {
1025 * It seems that the thread structures get zeroed out
1026 * at process exit. This makes it impossible to
1027 * support P_SETID, P_CID or P_CPUID.
1034 if (p_canwait(td, p)) {
1039 if (((options & WEXITED) == 0) && (p->p_state == PRS_ZOMBIE)) {
1045 * This special case handles a kthread spawned by linux_clone
1046 * (see linux_misc.c). The linux_wait4 and linux_waitpid
1047 * functions need to be able to distinguish between waiting
1048 * on a process and waiting on a thread. It is a thread if
1049 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option
1050 * signifies we want to wait for threads and not processes.
1052 if ((p->p_sigparent != SIGCHLD) ^
1053 ((options & WLINUXCLONE) != 0)) {
1058 if (siginfo != NULL) {
1059 bzero(siginfo, sizeof(*siginfo));
1060 siginfo->si_errno = 0;
1063 * SUSv4 requires that the si_signo value is always
1064 * SIGCHLD. Obey it despite the rfork(2) interface
1065 * allows to request other signal for child exit
1068 siginfo->si_signo = SIGCHLD;
1071 * This is still a rough estimate. We will fix the
1072 * cases TRAPPED, STOPPED, and CONTINUED later.
1074 if (WCOREDUMP(p->p_xsig)) {
1075 siginfo->si_code = CLD_DUMPED;
1076 siginfo->si_status = WTERMSIG(p->p_xsig);
1077 } else if (WIFSIGNALED(p->p_xsig)) {
1078 siginfo->si_code = CLD_KILLED;
1079 siginfo->si_status = WTERMSIG(p->p_xsig);
1081 siginfo->si_code = CLD_EXITED;
1082 siginfo->si_status = p->p_xexit;
1085 siginfo->si_pid = p->p_pid;
1086 siginfo->si_uid = p->p_ucred->cr_uid;
1089 * The si_addr field would be useful additional
1090 * detail, but apparently the PC value may be lost
1091 * when we reach this point. bzero() above sets
1092 * siginfo->si_addr to NULL.
1097 * There should be no reason to limit resources usage info to
1098 * exited processes only. A snapshot about any resources used
1099 * by a stopped process may be exactly what is needed.
1101 if (wrusage != NULL) {
1102 rup = &wrusage->wru_self;
1105 calcru(p, &rup->ru_utime, &rup->ru_stime);
1108 rup = &wrusage->wru_children;
1109 *rup = p->p_stats->p_cru;
1110 calccru(p, &rup->ru_utime, &rup->ru_stime);
1113 if (p->p_state == PRS_ZOMBIE && !check_only) {
1114 proc_reap(td, p, status, options);
1121 kern_wait(struct thread *td, pid_t pid, int *status, int options,
1122 struct rusage *rusage)
1124 struct __wrusage wru, *wrup;
1130 * Translate the special pid values into the (idtype, pid)
1131 * pair for kern_wait6. The WAIT_MYPGRP case is handled by
1132 * kern_wait6() on its own.
1134 if (pid == WAIT_ANY) {
1137 } else if (pid < 0) {
1151 * For backward compatibility we implicitly add flags WEXITED
1152 * and WTRAPPED here.
1154 options |= WEXITED | WTRAPPED;
1155 ret = kern_wait6(td, idtype, id, status, options, wrup, NULL);
1157 *rusage = wru.wru_self;
1162 report_alive_proc(struct thread *td, struct proc *p, siginfo_t *siginfo,
1163 int *status, int options, int si_code)
1167 PROC_LOCK_ASSERT(p, MA_OWNED);
1168 sx_assert(&proctree_lock, SA_XLOCKED);
1169 MPASS(si_code == CLD_TRAPPED || si_code == CLD_STOPPED ||
1170 si_code == CLD_CONTINUED);
1172 cont = si_code == CLD_CONTINUED;
1173 if ((options & WNOWAIT) == 0) {
1175 p->p_flag &= ~P_CONTINUED;
1177 p->p_flag |= P_WAITED;
1178 PROC_LOCK(td->td_proc);
1179 sigqueue_take(p->p_ksi);
1180 PROC_UNLOCK(td->td_proc);
1182 sx_xunlock(&proctree_lock);
1183 if (siginfo != NULL) {
1184 siginfo->si_code = si_code;
1185 siginfo->si_status = cont ? SIGCONT : p->p_xsig;
1188 *status = cont ? SIGCONT : W_STOPCODE(p->p_xsig);
1190 td->td_retval[0] = p->p_pid;
1194 kern_wait6(struct thread *td, idtype_t idtype, id_t id, int *status,
1195 int options, struct __wrusage *wrusage, siginfo_t *siginfo)
1199 int error, nfound, ret;
1202 AUDIT_ARG_VALUE((int)idtype); /* XXX - This is likely wrong! */
1203 AUDIT_ARG_PID((pid_t)id); /* XXX - This may be wrong! */
1204 AUDIT_ARG_VALUE(options);
1208 if ((pid_t)id == WAIT_MYPGRP && (idtype == P_PID || idtype == P_PGID)) {
1210 id = (id_t)q->p_pgid;
1215 /* If we don't know the option, just return. */
1216 if ((options & ~(WUNTRACED | WNOHANG | WCONTINUED | WNOWAIT |
1217 WEXITED | WTRAPPED | WLINUXCLONE)) != 0)
1219 if ((options & (WEXITED | WUNTRACED | WCONTINUED | WTRAPPED)) == 0) {
1221 * We will be unable to find any matching processes,
1222 * because there are no known events to look for.
1223 * Prefer to return error instead of blocking
1230 if (q->p_flag & P_STATCHILD) {
1232 q->p_flag &= ~P_STATCHILD;
1235 sx_xlock(&proctree_lock);
1238 LIST_FOREACH(p, &q->p_children, p_sibling) {
1240 ret = proc_to_reap(td, p, idtype, id, status, options,
1241 wrusage, siginfo, 0);
1244 else if (ret != 1) {
1245 td->td_retval[0] = pid;
1250 PROC_LOCK_ASSERT(p, MA_OWNED);
1252 if ((options & WTRAPPED) != 0 &&
1253 (p->p_flag & P_TRACED) != 0) {
1256 ((p->p_flag & (P_STOPPED_TRACE | P_STOPPED_SIG)) &&
1257 p->p_suspcount == p->p_numthreads &&
1258 (p->p_flag & P_WAITED) == 0);
1262 "wait: returning trapped pid %d status %#x "
1263 "(xstat %d) xthread %d",
1264 p->p_pid, W_STOPCODE(p->p_xsig), p->p_xsig,
1265 p->p_xthread != NULL ?
1266 p->p_xthread->td_tid : -1);
1267 report_alive_proc(td, p, siginfo, status,
1268 options, CLD_TRAPPED);
1272 if ((options & WUNTRACED) != 0 &&
1273 (p->p_flag & P_STOPPED_SIG) != 0) {
1275 report = (p->p_suspcount == p->p_numthreads &&
1276 ((p->p_flag & P_WAITED) == 0));
1279 report_alive_proc(td, p, siginfo, status,
1280 options, CLD_STOPPED);
1284 if ((options & WCONTINUED) != 0 &&
1285 (p->p_flag & P_CONTINUED) != 0) {
1286 report_alive_proc(td, p, siginfo, status, options,
1294 * Look in the orphans list too, to allow the parent to
1295 * collect it's child exit status even if child is being
1298 * Debugger detaches from the parent upon successful
1299 * switch-over from parent to child. At this point due to
1300 * re-parenting the parent loses the child to debugger and a
1301 * wait4(2) call would report that it has no children to wait
1302 * for. By maintaining a list of orphans we allow the parent
1303 * to successfully wait until the child becomes a zombie.
1306 LIST_FOREACH(p, &q->p_orphans, p_orphan) {
1307 ret = proc_to_reap(td, p, idtype, id, NULL, options,
1310 KASSERT(ret != -1, ("reaped an orphan (pid %d)",
1311 (int)td->td_retval[0]));
1319 sx_xunlock(&proctree_lock);
1322 if (options & WNOHANG) {
1323 sx_xunlock(&proctree_lock);
1324 td->td_retval[0] = 0;
1328 if (q->p_flag & P_STATCHILD) {
1329 q->p_flag &= ~P_STATCHILD;
1333 sx_xunlock(&proctree_lock);
1334 error = msleep(q, &q->p_mtx, PWAIT | PCATCH | PDROP, "wait", 0);
1341 proc_add_orphan(struct proc *child, struct proc *parent)
1344 sx_assert(&proctree_lock, SX_XLOCKED);
1345 KASSERT((child->p_flag & P_TRACED) != 0,
1346 ("proc_add_orphan: not traced"));
1348 if (LIST_EMPTY(&parent->p_orphans)) {
1349 child->p_treeflag |= P_TREE_FIRST_ORPHAN;
1350 LIST_INSERT_HEAD(&parent->p_orphans, child, p_orphan);
1352 LIST_INSERT_AFTER(LIST_FIRST(&parent->p_orphans),
1355 child->p_treeflag |= P_TREE_ORPHANED;
1359 * Make process 'parent' the new parent of process 'child'.
1360 * Must be called with an exclusive hold of proctree lock.
1363 proc_reparent(struct proc *child, struct proc *parent, bool set_oppid)
1366 sx_assert(&proctree_lock, SX_XLOCKED);
1367 PROC_LOCK_ASSERT(child, MA_OWNED);
1368 if (child->p_pptr == parent)
1371 PROC_LOCK(child->p_pptr);
1372 sigqueue_take(child->p_ksi);
1373 PROC_UNLOCK(child->p_pptr);
1374 LIST_REMOVE(child, p_sibling);
1375 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
1377 proc_clear_orphan(child);
1378 if ((child->p_flag & P_TRACED) != 0) {
1379 proc_add_orphan(child, child->p_pptr);
1382 child->p_pptr = parent;
1384 child->p_oppid = parent->p_pid;