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("$FreeBSD$");
43 #include "opt_ktrace.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>
52 #include <sys/malloc.h>
54 #include <sys/mutex.h>
56 #include <sys/procdesc.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/sysctl.h>
70 #include <sys/syslog.h>
71 #include <sys/ptrace.h>
72 #include <sys/acct.h> /* for acct_process() function prototype */
73 #include <sys/filedesc.h>
77 #include <sys/sysent.h>
78 #include <sys/timers.h>
79 #include <sys/umtxvar.h>
81 #include <sys/ktrace.h>
84 #include <security/audit/audit.h>
85 #include <security/mac/mac_framework.h>
88 #include <vm/vm_extern.h>
89 #include <vm/vm_param.h>
91 #include <vm/vm_map.h>
92 #include <vm/vm_page.h>
96 #include <sys/dtrace_bsd.h>
97 dtrace_execexit_func_t dtrace_fasttrap_exit;
100 SDT_PROVIDER_DECLARE(proc);
101 SDT_PROBE_DEFINE1(proc, , , exit, "int");
103 static int kern_kill_on_dbg_exit = 1;
104 SYSCTL_INT(_kern, OID_AUTO, kill_on_debugger_exit, CTLFLAG_RWTUN,
105 &kern_kill_on_dbg_exit, 0,
106 "Kill ptraced processes when debugger exits");
108 static bool kern_wait_dequeue_sigchld = 1;
109 SYSCTL_BOOL(_kern, OID_AUTO, wait_dequeue_sigchld, CTLFLAG_RWTUN,
110 &kern_wait_dequeue_sigchld, 0,
111 "Dequeue SIGCHLD on wait(2) for live process");
114 proc_realparent(struct proc *child)
116 struct proc *p, *parent;
118 sx_assert(&proctree_lock, SX_LOCKED);
119 if ((child->p_treeflag & P_TREE_ORPHANED) == 0)
120 return (child->p_pptr->p_pid == child->p_oppid ?
121 child->p_pptr : child->p_reaper);
122 for (p = child; (p->p_treeflag & P_TREE_FIRST_ORPHAN) == 0;) {
123 /* Cannot use LIST_PREV(), since the list head is not known. */
124 p = __containerof(p->p_orphan.le_prev, struct proc,
126 KASSERT((p->p_treeflag & P_TREE_ORPHANED) != 0,
127 ("missing P_ORPHAN %p", p));
129 parent = __containerof(p->p_orphan.le_prev, struct proc,
135 reaper_abandon_children(struct proc *p, bool exiting)
137 struct proc *p1, *p2, *ptmp;
139 sx_assert(&proctree_lock, SX_XLOCKED);
140 KASSERT(p != initproc, ("reaper_abandon_children for initproc"));
141 if ((p->p_treeflag & P_TREE_REAPER) == 0)
144 LIST_FOREACH_SAFE(p2, &p->p_reaplist, p_reapsibling, ptmp) {
145 LIST_REMOVE(p2, p_reapsibling);
147 p2->p_reapsubtree = p->p_reapsubtree;
148 LIST_INSERT_HEAD(&p1->p_reaplist, p2, p_reapsibling);
149 if (exiting && p2->p_pptr == p) {
151 proc_reparent(p2, p1, true);
155 KASSERT(LIST_EMPTY(&p->p_reaplist), ("p_reaplist not empty"));
156 p->p_treeflag &= ~P_TREE_REAPER;
160 reaper_clear(struct proc *p)
165 sx_assert(&proctree_lock, SX_LOCKED);
166 LIST_REMOVE(p, p_reapsibling);
167 if (p->p_reapsubtree == 1)
170 LIST_FOREACH(p1, &p->p_reaper->p_reaplist, p_reapsibling) {
171 if (p1->p_reapsubtree == p->p_reapsubtree) {
177 proc_id_clear(PROC_ID_REAP, p->p_reapsubtree);
181 proc_clear_orphan(struct proc *p)
185 sx_assert(&proctree_lock, SA_XLOCKED);
186 if ((p->p_treeflag & P_TREE_ORPHANED) == 0)
188 if ((p->p_treeflag & P_TREE_FIRST_ORPHAN) != 0) {
189 p1 = LIST_NEXT(p, p_orphan);
191 p1->p_treeflag |= P_TREE_FIRST_ORPHAN;
192 p->p_treeflag &= ~P_TREE_FIRST_ORPHAN;
194 LIST_REMOVE(p, p_orphan);
195 p->p_treeflag &= ~P_TREE_ORPHANED;
199 exit_onexit(struct proc *p)
201 MPASS(p->p_numthreads == 1);
202 umtx_thread_exit(FIRST_THREAD_IN_PROC(p));
206 * exit -- death of process.
209 sys_exit(struct thread *td, struct exit_args *uap)
212 exit1(td, uap->rval, 0);
217 proc_set_p2_wexit(struct proc *p)
219 PROC_LOCK_ASSERT(p, MA_OWNED);
220 p->p_flag2 |= P2_WEXIT;
224 * Exit: deallocate address space and other resources, change proc state to
225 * zombie, and unlink proc from allproc and parent's lists. Save exit status
226 * and rusage for wait(). Check for child processes and orphan them.
229 exit1(struct thread *td, int rval, int signo)
231 struct proc *p, *nq, *q, *t;
233 ksiginfo_t *ksi, *ksi1;
236 mtx_assert(&Giant, MA_NOTOWNED);
237 KASSERT(rval == 0 || signo == 0, ("exit1 rv %d sig %d", rval, signo));
238 TSPROCEXIT(td->td_proc->p_pid);
242 * XXX in case we're rebooting we just let init die in order to
243 * work around an unsolved stack overflow seen very late during
244 * shutdown on sparc64 when the gmirror worker process exists.
245 * XXX what to do now that sparc64 is gone... remove if?
247 if (p == initproc && rebooting == 0) {
248 printf("init died (signal %d, exit %d)\n", signo, rval);
249 panic("Going nowhere without my init!");
253 * Process deferred operations, designated with ASTF_KCLEAR.
254 * For instance, we need to deref SU mp, since the thread does
255 * not return to userspace, and wait for geom to stabilize.
260 * MUST abort all other threads before proceeding past here.
263 proc_set_p2_wexit(p);
266 * First check if some other thread or external request got
267 * here before us. If so, act appropriately: exit or suspend.
268 * We must ensure that stop requests are handled before we set
271 thread_suspend_check(0);
272 while (p->p_flag & P_HADTHREADS) {
274 * Kill off the other threads. This requires
275 * some co-operation from other parts of the kernel
276 * so it may not be instantaneous. With this state set
277 * any thread attempting to interruptibly
278 * sleep will return immediately with EINTR or EWOULDBLOCK
279 * which will hopefully force them to back out to userland
280 * freeing resources as they go. Any thread attempting
281 * to return to userland will thread_exit() from ast().
282 * thread_exit() will unsuspend us when the last of the
283 * other threads exits.
284 * If there is already a thread singler after resumption,
285 * calling thread_single() will fail; in that case, we just
286 * re-check all suspension request, the thread should
287 * either be suspended there or exit.
289 if (!thread_single(p, SINGLE_EXIT))
291 * All other activity in this process is now
292 * stopped. Threading support has been turned
297 * Recheck for new stop or suspend requests which
298 * might appear while process lock was dropped in
301 thread_suspend_check(0);
303 KASSERT(p->p_numthreads == 1,
304 ("exit1: proc %p exiting with %d threads", p, p->p_numthreads));
305 racct_sub(p, RACCT_NTHR, 1);
307 /* Let event handler change exit status */
312 * Ignore any pending request to stop due to a stop signal.
313 * Once P_WEXIT is set, future requests will be ignored as
316 p->p_flag &= ~P_STOPPED_SIG;
317 KASSERT(!P_SHOULDSTOP(p), ("exiting process is stopped"));
319 /* Note that we are exiting. */
320 p->p_flag |= P_WEXIT;
323 * Wait for any processes that have a hold on our vmspace to
324 * release their reference.
326 while (p->p_lock > 0)
327 msleep(&p->p_lock, &p->p_mtx, PWAIT, "exithold", 0);
330 /* Drain the limit callout while we don't have the proc locked */
331 callout_drain(&p->p_limco);
335 * The Sun BSM exit token contains two components: an exit status as
336 * passed to exit(), and a return value to indicate what sort of exit
337 * it was. The exit status is WEXITSTATUS(rv), but it's not clear
338 * what the return value is.
340 AUDIT_ARG_EXIT(rval, 0);
341 AUDIT_SYSCALL_EXIT(0, td);
344 /* Are we a task leader with peers? */
345 if (p->p_peers != NULL && p == p->p_leader) {
346 mtx_lock(&ppeers_lock);
350 kern_psignal(q, SIGKILL);
354 while (p->p_peers != NULL)
355 msleep(p, &ppeers_lock, PWAIT, "exit1", 0);
356 mtx_unlock(&ppeers_lock);
362 * Check if any loadable modules need anything done at process exit.
363 * E.g. SYSV IPC stuff.
364 * Event handler could change exit status.
365 * XXX what if one of these generates an error?
367 EVENTHANDLER_DIRECT_INVOKE(process_exit, p);
370 * If parent is waiting for us to exit or exec,
371 * P_PPWAIT is set; we will wakeup the parent below.
378 * Stop the real interval timer. If the handler is currently
379 * executing, prevent it from rearming itself and let it finish.
381 if (timevalisset(&p->p_realtimer.it_value) &&
382 callout_stop(&p->p_itcallout) == 0) {
383 timevalclear(&p->p_realtimer.it_interval);
385 callout_drain(&p->p_itcallout);
390 if (p->p_sysent->sv_onexit != NULL)
391 p->p_sysent->sv_onexit(p);
395 * Reset any sigio structures pointing to us as a result of
396 * F_SETOWN with our pid. The P_WEXIT flag interlocks with fsetown().
398 funsetownlst(&p->p_sigiolst);
401 * Close open files and release open-file table.
408 * Remove ourself from our leader's peer list and wake our leader.
410 if (p->p_leader->p_peers != NULL) {
411 mtx_lock(&ppeers_lock);
412 if (p->p_leader->p_peers != NULL) {
414 while (q->p_peers != p)
416 q->p_peers = p->p_peers;
419 mtx_unlock(&ppeers_lock);
422 exec_free_abi_mappings(p);
424 (void)acct_process(td);
430 * Release reference to text vnode etc
432 if (p->p_textvp != NULL) {
436 if (p->p_textdvp != NULL) {
440 if (p->p_binname != NULL) {
441 free(p->p_binname, M_PARGS);
446 * Release our limits structure.
448 lim_free(p->p_limit);
454 * Call machine-dependent code to release any
455 * machine-dependent resources other than the address space.
456 * The address space is released by "vmspace_exitfree(p)" in
461 WITNESS_WARN(WARN_PANIC, NULL, "process (pid %d) exiting", p->p_pid);
464 * Remove from allproc. It still sits in the hash.
466 sx_xlock(&allproc_lock);
467 LIST_REMOVE(p, p_list);
471 * Used by ddb's 'ps' command to find this process via the
474 p->p_list.le_prev = NULL;
476 prison_proc_unlink(p->p_ucred->cr_prison, p);
477 sx_xunlock(&allproc_lock);
479 sx_xlock(&proctree_lock);
480 if ((p->p_flag & (P_TRACED | P_PPWAIT | P_PPTRACE)) != 0) {
482 p->p_flag &= ~(P_TRACED | P_PPWAIT | P_PPTRACE);
487 * killjobc() might drop and re-acquire proctree_lock to
488 * revoke control tty if exiting process was a session leader.
493 * Reparent all children processes:
494 * - traced ones to the original parent (or init if we are that parent)
497 q = LIST_FIRST(&p->p_children);
498 if (q != NULL) /* only need this if any child is S_ZOMB */
500 for (; q != NULL; q = nq) {
501 nq = LIST_NEXT(q, p_sibling);
502 ksi = ksiginfo_alloc(M_WAITOK);
504 q->p_sigparent = SIGCHLD;
506 if ((q->p_flag & P_TRACED) == 0) {
507 proc_reparent(q, q->p_reaper, true);
508 if (q->p_state == PRS_ZOMBIE) {
510 * Inform reaper about the reparented
511 * zombie, since wait(2) has something
512 * new to report. Guarantee queueing
513 * of the SIGCHLD signal, similar to
514 * the _exit() behaviour, by providing
515 * our ksiginfo. Ksi is freed by the
518 if (q->p_ksi == NULL) {
521 ksiginfo_copy(q->p_ksi, ksi);
522 ksi->ksi_flags |= KSI_INS;
526 PROC_LOCK(q->p_reaper);
527 pksignal(q->p_reaper, SIGCHLD, ksi1);
528 PROC_UNLOCK(q->p_reaper);
529 } else if (q->p_pdeathsig > 0) {
531 * The child asked to received a signal
534 kern_psignal(q, q->p_pdeathsig);
538 * Traced processes are killed by default
539 * since their existence means someone is
542 t = proc_realparent(q);
544 proc_reparent(q, q->p_reaper, true);
547 proc_reparent(q, t, true);
551 * Since q was found on our children list, the
552 * proc_reparent() call moved q to the orphan
553 * list due to present P_TRACED flag. Clear
554 * orphan link for q now while q is locked.
556 proc_clear_orphan(q);
557 q->p_flag &= ~P_TRACED;
558 q->p_flag2 &= ~P2_PTRACE_FSTP;
561 FOREACH_THREAD_IN_PROC(q, tdt) {
562 tdt->td_dbgflags &= ~(TDB_SUSPEND | TDB_XSIG |
566 if (kern_kill_on_dbg_exit) {
567 q->p_flag &= ~P_STOPPED_TRACE;
568 kern_psignal(q, SIGKILL);
569 } else if ((q->p_flag & (P_STOPPED_TRACE |
570 P_STOPPED_SIG)) != 0) {
571 sigqueue_delete_proc(q, SIGTRAP);
581 * Also get rid of our orphans.
583 while ((q = LIST_FIRST(&p->p_orphans)) != NULL) {
585 KASSERT(q->p_oppid == p->p_pid,
586 ("orphan %p of %p has unexpected oppid %d", q, p,
588 q->p_oppid = q->p_reaper->p_pid;
591 * If we are the real parent of this process
592 * but it has been reparented to a debugger, then
593 * check if it asked for a signal when we exit.
595 if (q->p_pdeathsig > 0)
596 kern_psignal(q, q->p_pdeathsig);
597 CTR2(KTR_PTRACE, "exit: pid %d, clearing orphan %d", p->p_pid,
599 proc_clear_orphan(q);
604 if (SDT_PROBES_ENABLED()) {
605 int reason = CLD_EXITED;
606 if (WCOREDUMP(signo))
608 else if (WIFSIGNALED(signo))
610 SDT_PROBE1(proc, , , exit, reason);
614 /* Save exit status. */
618 if (p->p_sysent->sv_ontdexit != NULL)
619 p->p_sysent->sv_ontdexit(td);
623 * Tell the DTrace fasttrap provider about the exit if it
624 * has declared an interest.
626 if (dtrace_fasttrap_exit)
627 dtrace_fasttrap_exit(p);
631 * Notify interested parties of our demise.
633 KNOTE_LOCKED(p->p_klist, NOTE_EXIT);
636 * If this is a process with a descriptor, we may not need to deliver
637 * a signal to the parent. proctree_lock is held over
638 * procdesc_exit() to serialize concurrent calls to close() and
642 if (p->p_procdesc == NULL || procdesc_exit(p)) {
644 * Notify parent that we're gone. If parent has the
645 * PS_NOCLDWAIT flag set, or if the handler is set to SIG_IGN,
646 * notify process 1 instead (and hope it will handle this
649 PROC_LOCK(p->p_pptr);
650 mtx_lock(&p->p_pptr->p_sigacts->ps_mtx);
651 if (p->p_pptr->p_sigacts->ps_flag &
652 (PS_NOCLDWAIT | PS_CLDSIGIGN)) {
655 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
658 proc_reparent(p, p->p_reaper, true);
659 p->p_sigparent = SIGCHLD;
660 PROC_LOCK(p->p_pptr);
663 * Notify parent, so in case he was wait(2)ing or
664 * executing waitpid(2) with our pid, he will
669 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
671 if (p->p_pptr == p->p_reaper || p->p_pptr == initproc) {
673 } else if (p->p_sigparent != 0) {
674 if (p->p_sigparent == SIGCHLD) {
676 } else { /* LINUX thread */
681 PROC_LOCK(p->p_pptr);
682 sx_xunlock(&proctree_lock);
684 if (signal_parent == 1) {
686 } else if (signal_parent == 2) {
687 kern_psignal(p->p_pptr, p->p_sigparent);
690 /* Tell the prison that we are gone. */
691 prison_proc_free(p->p_ucred->cr_prison);
694 * The state PRS_ZOMBIE prevents other proesses from sending
695 * signal to the process, to avoid memory leak, we free memory
696 * for signal queue at the time when the state is set.
698 sigqueue_flush(&p->p_sigqueue);
699 sigqueue_flush(&td->td_sigqueue);
702 * We have to wait until after acquiring all locks before
703 * changing p_state. We need to avoid all possible context
704 * switches (including ones from blocking on a mutex) while
705 * marked as a zombie. We also have to set the zombie state
706 * before we release the parent process' proc lock to avoid
707 * a lost wakeup. So, we first call wakeup, then we grab the
708 * sched lock, update the state, and release the parent process'
712 cv_broadcast(&p->p_pwait);
713 sched_exit(p->p_pptr, td);
715 p->p_state = PRS_ZOMBIE;
716 PROC_UNLOCK(p->p_pptr);
719 * Save our children's rusage information in our exit rusage.
722 ruadd(&p->p_ru, &p->p_rux, &p->p_stats->p_cru, &p->p_crux);
726 * Make sure the scheduler takes this thread out of its tables etc.
727 * This will also release this thread's reference to the ucred.
728 * Other thread parts to release include pcb bits and such.
733 #ifndef _SYS_SYSPROTO_H_
742 sys_abort2(struct thread *td, struct abort2_args *uap)
749 if (nargs < 0 || nargs > nitems(uargs))
753 if (uap->args != NULL) {
754 error = copyin(uap->args, uargs,
755 nargs * sizeof(void *));
763 return (kern_abort2(td, uap->why, nargs, uargsp));
769 * why - user pointer to why
770 * nargs - number of arguments copied or -1 if an error occurred in copying
771 * args - pointer to an array of pointers in kernel format
774 kern_abort2(struct thread *td, const char *why, int nargs, void **uargs)
776 struct proc *p = td->td_proc;
781 * Do it right now so we can log either proper call of abort2(), or
782 * note, that invalid argument was passed. 512 is big enough to
783 * handle 16 arguments' descriptions with additional comments.
785 sb = sbuf_new(NULL, NULL, 512, SBUF_FIXEDLEN);
787 sbuf_printf(sb, "%s(pid %d uid %d) aborted: ",
788 p->p_comm, p->p_pid, td->td_ucred->cr_uid);
790 * Since we can't return from abort2(), send SIGKILL in cases, where
791 * abort2() was called improperly
794 /* Prevent from DoSes from user-space. */
797 KASSERT(nargs >= 0 && nargs <= 16, ("called with too many args (%d)",
800 * Limit size of 'reason' string to 128. Will fit even when
801 * maximal number of arguments was chosen to be logged.
804 error = sbuf_copyin(sb, why, 128);
808 sbuf_printf(sb, "(null)");
811 sbuf_printf(sb, "(");
812 for (i = 0;i < nargs; i++)
813 sbuf_printf(sb, "%s%p", i == 0 ? "" : ", ", uargs[i]);
814 sbuf_printf(sb, ")");
817 * Final stage: arguments were proper, string has been
818 * successfully copied from userspace, and copying pointers
819 * from user-space succeed.
823 if (sig == SIGKILL) {
825 sbuf_printf(sb, " (Reason text inaccessible)");
829 log(LOG_INFO, "%s", sbuf_data(sb));
837 * The dirty work is handled by kern_wait().
840 owait(struct thread *td, struct owait_args *uap __unused)
844 error = kern_wait(td, WAIT_ANY, &status, 0, NULL);
846 td->td_retval[1] = status;
849 #endif /* COMPAT_43 */
852 * The dirty work is handled by kern_wait().
855 sys_wait4(struct thread *td, struct wait4_args *uap)
857 struct rusage ru, *rup;
860 if (uap->rusage != NULL)
864 error = kern_wait(td, uap->pid, &status, uap->options, rup);
865 if (uap->status != NULL && error == 0 && td->td_retval[0] != 0)
866 error = copyout(&status, uap->status, sizeof(status));
867 if (uap->rusage != NULL && error == 0 && td->td_retval[0] != 0)
868 error = copyout(&ru, uap->rusage, sizeof(struct rusage));
873 sys_wait6(struct thread *td, struct wait6_args *uap)
875 struct __wrusage wru, *wrup;
881 idtype = uap->idtype;
884 if (uap->wrusage != NULL)
889 if (uap->info != NULL) {
891 bzero(sip, sizeof(*sip));
896 * We expect all callers of wait6() to know about WEXITED and
899 error = kern_wait6(td, idtype, id, &status, uap->options, wrup, sip);
901 if (uap->status != NULL && error == 0 && td->td_retval[0] != 0)
902 error = copyout(&status, uap->status, sizeof(status));
903 if (uap->wrusage != NULL && error == 0 && td->td_retval[0] != 0)
904 error = copyout(&wru, uap->wrusage, sizeof(wru));
905 if (uap->info != NULL && error == 0)
906 error = copyout(&si, uap->info, sizeof(si));
911 * Reap the remains of a zombie process and optionally return status and
912 * rusage. Asserts and will release both the proctree_lock and the process
913 * lock as part of its work.
916 proc_reap(struct thread *td, struct proc *p, int *status, int options)
920 sx_assert(&proctree_lock, SA_XLOCKED);
921 PROC_LOCK_ASSERT(p, MA_OWNED);
922 KASSERT(p->p_state == PRS_ZOMBIE, ("proc_reap: !PRS_ZOMBIE"));
924 mtx_spin_wait_unlocked(&p->p_slock);
929 *status = KW_EXITCODE(p->p_xexit, p->p_xsig);
930 if (options & WNOWAIT) {
932 * Only poll, returning the status. Caller does not wish to
933 * release the proc struct just yet.
936 sx_xunlock(&proctree_lock);
941 sigqueue_take(p->p_ksi);
945 * If we got the child via a ptrace 'attach', we need to give it back
948 if (p->p_oppid != p->p_pptr->p_pid) {
950 t = proc_realparent(p);
954 "wait: traced child %d moved back to parent %d", p->p_pid,
956 proc_reparent(p, t, false);
958 pksignal(t, SIGCHLD, p->p_ksi);
960 cv_broadcast(&p->p_pwait);
962 sx_xunlock(&proctree_lock);
968 * Remove other references to this process to ensure we have an
969 * exclusive reference.
971 sx_xlock(PIDHASHLOCK(p->p_pid));
972 LIST_REMOVE(p, p_hash);
973 sx_xunlock(PIDHASHLOCK(p->p_pid));
974 LIST_REMOVE(p, p_sibling);
975 reaper_abandon_children(p, true);
978 proc_clear_orphan(p);
981 if (p->p_procdesc != NULL)
983 sx_xunlock(&proctree_lock);
985 proc_id_clear(PROC_ID_PID, p->p_pid);
988 knlist_detach(p->p_klist);
993 * Removal from allproc list and process group list paired with
994 * PROC_LOCK which was executed during that time should guarantee
995 * nothing can reach this process anymore. As such further locking
998 p->p_xexit = p->p_xsig = 0; /* XXX: why? */
1001 ruadd(&q->p_stats->p_cru, &q->p_crux, &p->p_ru, &p->p_rux);
1005 * Decrement the count of procs running with this uid.
1007 (void)chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0);
1010 * Destroy resource accounting information associated with the process.
1015 racct_sub(p, RACCT_NPROC, 1);
1022 * Free credentials, arguments, and sigacts.
1025 pargs_drop(p->p_args);
1027 sigacts_free(p->p_sigacts);
1028 p->p_sigacts = NULL;
1031 * Do any thread-system specific cleanups.
1036 * Give vm and machine-dependent layer a chance to free anything that
1037 * cpu_exit couldn't release while still running in process context.
1041 mac_proc_destroy(p);
1044 KASSERT(FIRST_THREAD_IN_PROC(p),
1045 ("proc_reap: no residual thread!"));
1046 uma_zfree(proc_zone, p);
1047 atomic_add_int(&nprocs, -1);
1051 proc_to_reap(struct thread *td, struct proc *p, idtype_t idtype, id_t id,
1052 int *status, int options, struct __wrusage *wrusage, siginfo_t *siginfo,
1057 sx_assert(&proctree_lock, SA_XLOCKED);
1063 if (p->p_procdesc == NULL ||
1064 (p->p_pptr == td->td_proc &&
1065 (p->p_flag & P_TRACED) != 0)) {
1072 if (p->p_pid != (pid_t)id) {
1078 if (p->p_pgid != (pid_t)id) {
1084 if (p->p_session->s_sid != (pid_t)id) {
1090 if (p->p_ucred->cr_uid != (uid_t)id) {
1096 if (p->p_ucred->cr_gid != (gid_t)id) {
1102 if (p->p_ucred->cr_prison->pr_id != (int)id) {
1108 * It seems that the thread structures get zeroed out
1109 * at process exit. This makes it impossible to
1110 * support P_SETID, P_CID or P_CPUID.
1117 if (p_canwait(td, p)) {
1122 if (((options & WEXITED) == 0) && (p->p_state == PRS_ZOMBIE)) {
1128 * This special case handles a kthread spawned by linux_clone
1129 * (see linux_misc.c). The linux_wait4 and linux_waitpid
1130 * functions need to be able to distinguish between waiting
1131 * on a process and waiting on a thread. It is a thread if
1132 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option
1133 * signifies we want to wait for threads and not processes.
1135 if ((p->p_sigparent != SIGCHLD) ^
1136 ((options & WLINUXCLONE) != 0)) {
1141 if (siginfo != NULL) {
1142 bzero(siginfo, sizeof(*siginfo));
1143 siginfo->si_errno = 0;
1146 * SUSv4 requires that the si_signo value is always
1147 * SIGCHLD. Obey it despite the rfork(2) interface
1148 * allows to request other signal for child exit
1151 siginfo->si_signo = SIGCHLD;
1154 * This is still a rough estimate. We will fix the
1155 * cases TRAPPED, STOPPED, and CONTINUED later.
1157 if (WCOREDUMP(p->p_xsig)) {
1158 siginfo->si_code = CLD_DUMPED;
1159 siginfo->si_status = WTERMSIG(p->p_xsig);
1160 } else if (WIFSIGNALED(p->p_xsig)) {
1161 siginfo->si_code = CLD_KILLED;
1162 siginfo->si_status = WTERMSIG(p->p_xsig);
1164 siginfo->si_code = CLD_EXITED;
1165 siginfo->si_status = p->p_xexit;
1168 siginfo->si_pid = p->p_pid;
1169 siginfo->si_uid = p->p_ucred->cr_uid;
1172 * The si_addr field would be useful additional
1173 * detail, but apparently the PC value may be lost
1174 * when we reach this point. bzero() above sets
1175 * siginfo->si_addr to NULL.
1180 * There should be no reason to limit resources usage info to
1181 * exited processes only. A snapshot about any resources used
1182 * by a stopped process may be exactly what is needed.
1184 if (wrusage != NULL) {
1185 rup = &wrusage->wru_self;
1188 calcru(p, &rup->ru_utime, &rup->ru_stime);
1191 rup = &wrusage->wru_children;
1192 *rup = p->p_stats->p_cru;
1193 calccru(p, &rup->ru_utime, &rup->ru_stime);
1196 if (p->p_state == PRS_ZOMBIE && !check_only) {
1197 proc_reap(td, p, status, options);
1204 kern_wait(struct thread *td, pid_t pid, int *status, int options,
1205 struct rusage *rusage)
1207 struct __wrusage wru, *wrup;
1213 * Translate the special pid values into the (idtype, pid)
1214 * pair for kern_wait6. The WAIT_MYPGRP case is handled by
1215 * kern_wait6() on its own.
1217 if (pid == WAIT_ANY) {
1220 } else if (pid < 0) {
1234 * For backward compatibility we implicitly add flags WEXITED
1235 * and WTRAPPED here.
1237 options |= WEXITED | WTRAPPED;
1238 ret = kern_wait6(td, idtype, id, status, options, wrup, NULL);
1240 *rusage = wru.wru_self;
1245 report_alive_proc(struct thread *td, struct proc *p, siginfo_t *siginfo,
1246 int *status, int options, int si_code)
1250 PROC_LOCK_ASSERT(p, MA_OWNED);
1251 sx_assert(&proctree_lock, SA_XLOCKED);
1252 MPASS(si_code == CLD_TRAPPED || si_code == CLD_STOPPED ||
1253 si_code == CLD_CONTINUED);
1255 cont = si_code == CLD_CONTINUED;
1256 if ((options & WNOWAIT) == 0) {
1258 p->p_flag &= ~P_CONTINUED;
1260 p->p_flag |= P_WAITED;
1261 if (kern_wait_dequeue_sigchld &&
1262 (td->td_proc->p_sysent->sv_flags & SV_SIG_WAITNDQ) == 0) {
1263 PROC_LOCK(td->td_proc);
1264 sigqueue_take(p->p_ksi);
1265 PROC_UNLOCK(td->td_proc);
1268 sx_xunlock(&proctree_lock);
1269 if (siginfo != NULL) {
1270 siginfo->si_code = si_code;
1271 siginfo->si_status = cont ? SIGCONT : p->p_xsig;
1274 *status = cont ? SIGCONT : W_STOPCODE(p->p_xsig);
1275 td->td_retval[0] = p->p_pid;
1280 kern_wait6(struct thread *td, idtype_t idtype, id_t id, int *status,
1281 int options, struct __wrusage *wrusage, siginfo_t *siginfo)
1285 int error, nfound, ret;
1288 AUDIT_ARG_VALUE((int)idtype); /* XXX - This is likely wrong! */
1289 AUDIT_ARG_PID((pid_t)id); /* XXX - This may be wrong! */
1290 AUDIT_ARG_VALUE(options);
1294 if ((pid_t)id == WAIT_MYPGRP && (idtype == P_PID || idtype == P_PGID)) {
1296 id = (id_t)q->p_pgid;
1301 /* If we don't know the option, just return. */
1302 if ((options & ~(WUNTRACED | WNOHANG | WCONTINUED | WNOWAIT |
1303 WEXITED | WTRAPPED | WLINUXCLONE)) != 0)
1305 if ((options & (WEXITED | WUNTRACED | WCONTINUED | WTRAPPED)) == 0) {
1307 * We will be unable to find any matching processes,
1308 * because there are no known events to look for.
1309 * Prefer to return error instead of blocking
1316 if (q->p_flag & P_STATCHILD) {
1318 q->p_flag &= ~P_STATCHILD;
1321 sx_xlock(&proctree_lock);
1324 LIST_FOREACH(p, &q->p_children, p_sibling) {
1326 ret = proc_to_reap(td, p, idtype, id, status, options,
1327 wrusage, siginfo, 0);
1330 else if (ret != 1) {
1331 td->td_retval[0] = pid;
1336 PROC_LOCK_ASSERT(p, MA_OWNED);
1338 if ((options & WTRAPPED) != 0 &&
1339 (p->p_flag & P_TRACED) != 0) {
1342 ((p->p_flag & (P_STOPPED_TRACE | P_STOPPED_SIG)) &&
1343 p->p_suspcount == p->p_numthreads &&
1344 (p->p_flag & P_WAITED) == 0);
1348 "wait: returning trapped pid %d status %#x "
1349 "(xstat %d) xthread %d",
1350 p->p_pid, W_STOPCODE(p->p_xsig), p->p_xsig,
1351 p->p_xthread != NULL ?
1352 p->p_xthread->td_tid : -1);
1353 report_alive_proc(td, p, siginfo, status,
1354 options, CLD_TRAPPED);
1358 if ((options & WUNTRACED) != 0 &&
1359 (p->p_flag & P_STOPPED_SIG) != 0) {
1361 report = (p->p_suspcount == p->p_numthreads &&
1362 ((p->p_flag & P_WAITED) == 0));
1365 report_alive_proc(td, p, siginfo, status,
1366 options, CLD_STOPPED);
1370 if ((options & WCONTINUED) != 0 &&
1371 (p->p_flag & P_CONTINUED) != 0) {
1372 report_alive_proc(td, p, siginfo, status, options,
1380 * Look in the orphans list too, to allow the parent to
1381 * collect it's child exit status even if child is being
1384 * Debugger detaches from the parent upon successful
1385 * switch-over from parent to child. At this point due to
1386 * re-parenting the parent loses the child to debugger and a
1387 * wait4(2) call would report that it has no children to wait
1388 * for. By maintaining a list of orphans we allow the parent
1389 * to successfully wait until the child becomes a zombie.
1392 LIST_FOREACH(p, &q->p_orphans, p_orphan) {
1393 ret = proc_to_reap(td, p, idtype, id, NULL, options,
1396 KASSERT(ret != -1, ("reaped an orphan (pid %d)",
1397 (int)td->td_retval[0]));
1405 sx_xunlock(&proctree_lock);
1408 if (options & WNOHANG) {
1409 sx_xunlock(&proctree_lock);
1410 td->td_retval[0] = 0;
1414 if (q->p_flag & P_STATCHILD) {
1415 q->p_flag &= ~P_STATCHILD;
1419 sx_xunlock(&proctree_lock);
1420 error = msleep(q, &q->p_mtx, PWAIT | PCATCH | PDROP, "wait", 0);
1427 proc_add_orphan(struct proc *child, struct proc *parent)
1430 sx_assert(&proctree_lock, SX_XLOCKED);
1431 KASSERT((child->p_flag & P_TRACED) != 0,
1432 ("proc_add_orphan: not traced"));
1434 if (LIST_EMPTY(&parent->p_orphans)) {
1435 child->p_treeflag |= P_TREE_FIRST_ORPHAN;
1436 LIST_INSERT_HEAD(&parent->p_orphans, child, p_orphan);
1438 LIST_INSERT_AFTER(LIST_FIRST(&parent->p_orphans),
1441 child->p_treeflag |= P_TREE_ORPHANED;
1445 * Make process 'parent' the new parent of process 'child'.
1446 * Must be called with an exclusive hold of proctree lock.
1449 proc_reparent(struct proc *child, struct proc *parent, bool set_oppid)
1452 sx_assert(&proctree_lock, SX_XLOCKED);
1453 PROC_LOCK_ASSERT(child, MA_OWNED);
1454 if (child->p_pptr == parent)
1457 PROC_LOCK(child->p_pptr);
1458 sigqueue_take(child->p_ksi);
1459 PROC_UNLOCK(child->p_pptr);
1460 LIST_REMOVE(child, p_sibling);
1461 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
1463 proc_clear_orphan(child);
1464 if ((child->p_flag & P_TRACED) != 0) {
1465 proc_add_orphan(child, child->p_pptr);
1468 child->p_pptr = parent;
1470 child->p_oppid = parent->p_pid;