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>
50 #include <sys/malloc.h>
52 #include <sys/mutex.h>
54 #include <sys/procdesc.h>
55 #include <sys/pioctl.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>
77 #include <sys/ktrace.h>
80 #include <security/audit/audit.h>
81 #include <security/mac/mac_framework.h>
84 #include <vm/vm_extern.h>
85 #include <vm/vm_param.h>
87 #include <vm/vm_map.h>
88 #include <vm/vm_page.h>
92 #include <sys/dtrace_bsd.h>
93 dtrace_execexit_func_t dtrace_fasttrap_exit;
96 SDT_PROVIDER_DECLARE(proc);
97 SDT_PROBE_DEFINE1(proc, , , exit, "int");
99 /* Hook for NFS teardown procedure. */
100 void (*nlminfo_release_p)(struct proc *p);
102 EVENTHANDLER_LIST_DECLARE(process_exit);
105 proc_realparent(struct proc *child)
107 struct proc *p, *parent;
109 sx_assert(&proctree_lock, SX_LOCKED);
110 if ((child->p_treeflag & P_TREE_ORPHANED) == 0)
111 return (child->p_pptr->p_pid == child->p_oppid ?
112 child->p_pptr : initproc);
113 for (p = child; (p->p_treeflag & P_TREE_FIRST_ORPHAN) == 0;) {
114 /* Cannot use LIST_PREV(), since the list head is not known. */
115 p = __containerof(p->p_orphan.le_prev, struct proc,
117 KASSERT((p->p_treeflag & P_TREE_ORPHANED) != 0,
118 ("missing P_ORPHAN %p", p));
120 parent = __containerof(p->p_orphan.le_prev, struct proc,
126 reaper_abandon_children(struct proc *p, bool exiting)
128 struct proc *p1, *p2, *ptmp;
130 sx_assert(&proctree_lock, SX_LOCKED);
131 KASSERT(p != initproc, ("reaper_abandon_children for initproc"));
132 if ((p->p_treeflag & P_TREE_REAPER) == 0)
135 LIST_FOREACH_SAFE(p2, &p->p_reaplist, p_reapsibling, ptmp) {
136 LIST_REMOVE(p2, p_reapsibling);
138 p2->p_reapsubtree = p->p_reapsubtree;
139 LIST_INSERT_HEAD(&p1->p_reaplist, p2, p_reapsibling);
140 if (exiting && p2->p_pptr == p) {
142 proc_reparent(p2, p1, true);
146 KASSERT(LIST_EMPTY(&p->p_reaplist), ("p_reaplist not empty"));
147 p->p_treeflag &= ~P_TREE_REAPER;
151 proc_clear_orphan(struct proc *p)
155 sx_assert(&proctree_lock, SA_XLOCKED);
156 if ((p->p_treeflag & P_TREE_ORPHANED) == 0)
158 if ((p->p_treeflag & P_TREE_FIRST_ORPHAN) != 0) {
159 p1 = LIST_NEXT(p, p_orphan);
161 p1->p_treeflag |= P_TREE_FIRST_ORPHAN;
162 p->p_treeflag &= ~P_TREE_FIRST_ORPHAN;
164 LIST_REMOVE(p, p_orphan);
165 p->p_treeflag &= ~P_TREE_ORPHANED;
169 * exit -- death of process.
172 sys_sys_exit(struct thread *td, struct sys_exit_args *uap)
175 exit1(td, uap->rval, 0);
180 * Exit: deallocate address space and other resources, change proc state to
181 * zombie, and unlink proc from allproc and parent's lists. Save exit status
182 * and rusage for wait(). Check for child processes and orphan them.
185 exit1(struct thread *td, int rval, int signo)
187 struct proc *p, *nq, *q, *t;
189 ksiginfo_t *ksi, *ksi1;
192 mtx_assert(&Giant, MA_NOTOWNED);
193 KASSERT(rval == 0 || signo == 0, ("exit1 rv %d sig %d", rval, signo));
197 * XXX in case we're rebooting we just let init die in order to
198 * work around an unsolved stack overflow seen very late during
199 * shutdown on sparc64 when the gmirror worker process exists.
201 if (p == initproc && rebooting == 0) {
202 printf("init died (signal %d, exit %d)\n", signo, rval);
203 panic("Going nowhere without my init!");
207 * Deref SU mp, since the thread does not return to userspace.
209 td_softdep_cleanup(td);
212 * MUST abort all other threads before proceeding past here.
216 * First check if some other thread or external request got
217 * here before us. If so, act appropriately: exit or suspend.
218 * We must ensure that stop requests are handled before we set
221 thread_suspend_check(0);
222 while (p->p_flag & P_HADTHREADS) {
224 * Kill off the other threads. This requires
225 * some co-operation from other parts of the kernel
226 * so it may not be instantaneous. With this state set
227 * any thread entering the kernel from userspace will
228 * thread_exit() in trap(). Any thread attempting to
229 * sleep will return immediately with EINTR or EWOULDBLOCK
230 * which will hopefully force them to back out to userland
231 * freeing resources as they go. Any thread attempting
232 * to return to userland will thread_exit() from userret().
233 * thread_exit() will unsuspend us when the last of the
234 * other threads exits.
235 * If there is already a thread singler after resumption,
236 * calling thread_single will fail; in that case, we just
237 * re-check all suspension request, the thread should
238 * either be suspended there or exit.
240 if (!thread_single(p, SINGLE_EXIT))
242 * All other activity in this process is now
243 * stopped. Threading support has been turned
248 * Recheck for new stop or suspend requests which
249 * might appear while process lock was dropped in
252 thread_suspend_check(0);
254 KASSERT(p->p_numthreads == 1,
255 ("exit1: proc %p exiting with %d threads", p, p->p_numthreads));
256 racct_sub(p, RACCT_NTHR, 1);
258 /* Let event handler change exit status */
263 * Wakeup anyone in procfs' PIOCWAIT. They should have a hold
264 * on our vmspace, so we should block below until they have
265 * released their reference to us. Note that if they have
266 * requested S_EXIT stops we will block here until they ack
269 _STOPEVENT(p, S_EXIT, 0);
272 * Ignore any pending request to stop due to a stop signal.
273 * Once P_WEXIT is set, future requests will be ignored as
276 p->p_flag &= ~P_STOPPED_SIG;
277 KASSERT(!P_SHOULDSTOP(p), ("exiting process is stopped"));
280 * Note that we are exiting and do another wakeup of anyone in
281 * PIOCWAIT in case they aren't listening for S_EXIT stops or
282 * decided to wait again after we told them we are exiting.
284 p->p_flag |= P_WEXIT;
288 * Wait for any processes that have a hold on our vmspace to
289 * release their reference.
291 while (p->p_lock > 0)
292 msleep(&p->p_lock, &p->p_mtx, PWAIT, "exithold", 0);
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(rval, 0);
306 AUDIT_SYSCALL_EXIT(0, td);
309 /* Are we a task leader with peers? */
310 if (p->p_peers != NULL && 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 * Event handler could change exit status.
328 * XXX what if one of these generates an error?
330 EVENTHANDLER_DIRECT_INVOKE(process_exit, p);
333 * If parent is waiting for us to exit or exec,
334 * P_PPWAIT is set; we will wakeup the parent below.
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_safe(&p->p_itcallout, CS_EXECUTING, NULL) == 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 umtx_thread_exit(td);
357 * Reset any sigio structures pointing to us as a result of
358 * F_SETOWN with our pid.
360 funsetownlst(&p->p_sigiolst);
363 * If this process has an nlminfo data area (for lockd), release it
365 if (nlminfo_release_p != NULL && p->p_nlminfo != NULL)
366 (*nlminfo_release_p)(p);
369 * Close open files and release open-file table.
375 * If this thread tickled GEOM, we need to wait for the giggling to
376 * stop before we return to userland
378 if (td->td_pflags & TDP_GEOM)
382 * Remove ourself from our leader's peer list and wake our leader.
384 if (p->p_leader->p_peers != NULL) {
385 mtx_lock(&ppeers_lock);
386 if (p->p_leader->p_peers != NULL) {
388 while (q->p_peers != p)
390 q->p_peers = p->p_peers;
393 mtx_unlock(&ppeers_lock);
398 (void)acct_process(td);
404 * Release reference to text vnode
406 if (p->p_textvp != NULL) {
412 * Release our limits structure.
414 lim_free(p->p_limit);
420 * Call machine-dependent code to release any
421 * machine-dependent resources other than the address space.
422 * The address space is released by "vmspace_exitfree(p)" in
427 WITNESS_WARN(WARN_PANIC, NULL, "process (pid %d) exiting", p->p_pid);
429 sx_xlock(&proctree_lock);
431 * Remove proc from allproc queue and pidhash chain.
432 * Place onto zombproc. Unlink from parent's child list.
434 sx_xlock(&allproc_lock);
435 LIST_REMOVE(p, p_list);
436 LIST_INSERT_HEAD(&zombproc, p, p_list);
437 LIST_REMOVE(p, p_hash);
438 sx_xunlock(&allproc_lock);
441 p->p_flag &= ~(P_TRACED | P_PPWAIT | P_PPTRACE);
445 * Reparent all children processes:
446 * - traced ones to the original parent (or init if we are that parent)
449 q = LIST_FIRST(&p->p_children);
450 if (q != NULL) /* only need this if any child is S_ZOMB */
452 for (; q != NULL; q = nq) {
453 nq = LIST_NEXT(q, p_sibling);
454 ksi = ksiginfo_alloc(TRUE);
456 q->p_sigparent = SIGCHLD;
458 if ((q->p_flag & P_TRACED) == 0) {
459 proc_reparent(q, q->p_reaper, true);
460 if (q->p_state == PRS_ZOMBIE) {
462 * Inform reaper about the reparented
463 * zombie, since wait(2) has something
464 * new to report. Guarantee queueing
465 * of the SIGCHLD signal, similar to
466 * the _exit() behaviour, by providing
467 * our ksiginfo. Ksi is freed by the
470 if (q->p_ksi == NULL) {
473 ksiginfo_copy(q->p_ksi, ksi);
474 ksi->ksi_flags |= KSI_INS;
478 PROC_LOCK(q->p_reaper);
479 pksignal(q->p_reaper, SIGCHLD, ksi1);
480 PROC_UNLOCK(q->p_reaper);
481 } else if (q->p_pdeathsig > 0) {
483 * The child asked to received a signal
486 kern_psignal(q, q->p_pdeathsig);
490 * Traced processes are killed since their existence
491 * means someone is screwing up.
493 t = proc_realparent(q);
495 proc_reparent(q, q->p_reaper, true);
498 proc_reparent(q, t, true);
502 * Since q was found on our children list, the
503 * proc_reparent() call moved q to the orphan
504 * list due to present P_TRACED flag. Clear
505 * orphan link for q now while q is locked.
507 proc_clear_orphan(q);
508 q->p_flag &= ~(P_TRACED | P_STOPPED_TRACE);
509 q->p_flag2 &= ~P2_PTRACE_FSTP;
511 FOREACH_THREAD_IN_PROC(q, tdt) {
512 tdt->td_dbgflags &= ~(TDB_SUSPEND | TDB_XSIG |
515 kern_psignal(q, SIGKILL);
523 * Also get rid of our orphans.
525 while ((q = LIST_FIRST(&p->p_orphans)) != NULL) {
527 KASSERT(q->p_oppid == p->p_pid,
528 ("orphan %p of %p has unexpected oppid %d", q, p,
530 q->p_oppid = q->p_reaper->p_pid;
533 * If we are the real parent of this process
534 * but it has been reparented to a debugger, then
535 * check if it asked for a signal when we exit.
537 if (q->p_pdeathsig > 0)
538 kern_psignal(q, q->p_pdeathsig);
539 CTR2(KTR_PTRACE, "exit: pid %d, clearing orphan %d", p->p_pid,
541 proc_clear_orphan(q);
545 /* Save exit status. */
551 * Tell the DTrace fasttrap provider about the exit if it
552 * has declared an interest.
554 if (dtrace_fasttrap_exit)
555 dtrace_fasttrap_exit(p);
559 * Notify interested parties of our demise.
561 KNOTE_LOCKED(p->p_klist, NOTE_EXIT);
564 int reason = CLD_EXITED;
565 if (WCOREDUMP(signo))
567 else if (WIFSIGNALED(signo))
569 SDT_PROBE1(proc, , , exit, reason);
573 * If this is a process with a descriptor, we may not need to deliver
574 * a signal to the parent. proctree_lock is held over
575 * procdesc_exit() to serialize concurrent calls to close() and
579 if (p->p_procdesc == NULL || procdesc_exit(p)) {
581 * Notify parent that we're gone. If parent has the
582 * PS_NOCLDWAIT flag set, or if the handler is set to SIG_IGN,
583 * notify process 1 instead (and hope it will handle this
586 PROC_LOCK(p->p_pptr);
587 mtx_lock(&p->p_pptr->p_sigacts->ps_mtx);
588 if (p->p_pptr->p_sigacts->ps_flag &
589 (PS_NOCLDWAIT | PS_CLDSIGIGN)) {
592 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
595 proc_reparent(p, p->p_reaper, true);
596 p->p_sigparent = SIGCHLD;
597 PROC_LOCK(p->p_pptr);
600 * Notify parent, so in case he was wait(2)ing or
601 * executing waitpid(2) with our pid, he will
606 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
608 if (p->p_pptr == p->p_reaper || p->p_pptr == initproc) {
610 } else if (p->p_sigparent != 0) {
611 if (p->p_sigparent == SIGCHLD) {
613 } else { /* LINUX thread */
618 PROC_LOCK(p->p_pptr);
619 sx_xunlock(&proctree_lock);
621 if (signal_parent == 1) {
623 } else if (signal_parent == 2) {
624 kern_psignal(p->p_pptr, p->p_sigparent);
627 /* Tell the prison that we are gone. */
628 prison_proc_free(p->p_ucred->cr_prison);
631 * The state PRS_ZOMBIE prevents other proesses from sending
632 * signal to the process, to avoid memory leak, we free memory
633 * for signal queue at the time when the state is set.
635 sigqueue_flush(&p->p_sigqueue);
636 sigqueue_flush(&td->td_sigqueue);
639 * We have to wait until after acquiring all locks before
640 * changing p_state. We need to avoid all possible context
641 * switches (including ones from blocking on a mutex) while
642 * marked as a zombie. We also have to set the zombie state
643 * before we release the parent process' proc lock to avoid
644 * a lost wakeup. So, we first call wakeup, then we grab the
645 * sched lock, update the state, and release the parent process'
649 cv_broadcast(&p->p_pwait);
650 sched_exit(p->p_pptr, td);
652 p->p_state = PRS_ZOMBIE;
653 PROC_UNLOCK(p->p_pptr);
656 * Save our children's rusage information in our exit rusage.
659 ruadd(&p->p_ru, &p->p_rux, &p->p_stats->p_cru, &p->p_crux);
663 * Make sure the scheduler takes this thread out of its tables etc.
664 * This will also release this thread's reference to the ucred.
665 * Other thread parts to release include pcb bits and such.
671 #ifndef _SYS_SYSPROTO_H_
680 sys_abort2(struct thread *td, struct abort2_args *uap)
682 struct proc *p = td->td_proc;
688 * Do it right now so we can log either proper call of abort2(), or
689 * note, that invalid argument was passed. 512 is big enough to
690 * handle 16 arguments' descriptions with additional comments.
692 sb = sbuf_new(NULL, NULL, 512, SBUF_FIXEDLEN);
694 sbuf_printf(sb, "%s(pid %d uid %d) aborted: ",
695 p->p_comm, p->p_pid, td->td_ucred->cr_uid);
697 * Since we can't return from abort2(), send SIGKILL in cases, where
698 * abort2() was called improperly
701 /* Prevent from DoSes from user-space. */
702 if (uap->nargs < 0 || uap->nargs > 16)
704 if (uap->nargs > 0) {
705 if (uap->args == NULL)
707 error = copyin(uap->args, uargs, uap->nargs * sizeof(void *));
712 * Limit size of 'reason' string to 128. Will fit even when
713 * maximal number of arguments was chosen to be logged.
715 if (uap->why != NULL) {
716 error = sbuf_copyin(sb, uap->why, 128);
720 sbuf_printf(sb, "(null)");
722 if (uap->nargs > 0) {
723 sbuf_printf(sb, "(");
724 for (i = 0;i < uap->nargs; i++)
725 sbuf_printf(sb, "%s%p", i == 0 ? "" : ", ", uargs[i]);
726 sbuf_printf(sb, ")");
729 * Final stage: arguments were proper, string has been
730 * successfully copied from userspace, and copying pointers
731 * from user-space succeed.
735 if (sig == SIGKILL) {
737 sbuf_printf(sb, " (Reason text inaccessible)");
741 log(LOG_INFO, "%s", sbuf_data(sb));
750 * The dirty work is handled by kern_wait().
753 owait(struct thread *td, struct owait_args *uap __unused)
757 error = kern_wait(td, WAIT_ANY, &status, 0, NULL);
759 td->td_retval[1] = status;
762 #endif /* COMPAT_43 */
765 * The dirty work is handled by kern_wait().
768 sys_wait4(struct thread *td, struct wait4_args *uap)
770 struct rusage ru, *rup;
773 if (uap->rusage != NULL)
777 error = kern_wait(td, uap->pid, &status, uap->options, rup);
778 if (uap->status != NULL && error == 0 && td->td_retval[0] != 0)
779 error = copyout(&status, uap->status, sizeof(status));
780 if (uap->rusage != NULL && error == 0 && td->td_retval[0] != 0)
781 error = copyout(&ru, uap->rusage, sizeof(struct rusage));
786 sys_wait6(struct thread *td, struct wait6_args *uap)
788 struct __wrusage wru, *wrup;
794 idtype = uap->idtype;
797 if (uap->wrusage != NULL)
802 if (uap->info != NULL) {
804 bzero(sip, sizeof(*sip));
809 * We expect all callers of wait6() to know about WEXITED and
812 error = kern_wait6(td, idtype, id, &status, uap->options, wrup, sip);
814 if (uap->status != NULL && error == 0 && td->td_retval[0] != 0)
815 error = copyout(&status, uap->status, sizeof(status));
816 if (uap->wrusage != NULL && error == 0 && td->td_retval[0] != 0)
817 error = copyout(&wru, uap->wrusage, sizeof(wru));
818 if (uap->info != NULL && error == 0)
819 error = copyout(&si, uap->info, sizeof(si));
824 * Reap the remains of a zombie process and optionally return status and
825 * rusage. Asserts and will release both the proctree_lock and the process
826 * lock as part of its work.
829 proc_reap(struct thread *td, struct proc *p, int *status, int options)
833 sx_assert(&proctree_lock, SA_XLOCKED);
834 PROC_LOCK_ASSERT(p, MA_OWNED);
835 KASSERT(p->p_state == PRS_ZOMBIE, ("proc_reap: !PRS_ZOMBIE"));
837 mtx_spin_wait_unlocked(&p->p_slock);
842 *status = KW_EXITCODE(p->p_xexit, p->p_xsig);
843 if (options & WNOWAIT) {
845 * Only poll, returning the status. Caller does not wish to
846 * release the proc struct just yet.
849 sx_xunlock(&proctree_lock);
854 sigqueue_take(p->p_ksi);
858 * If we got the child via a ptrace 'attach', we need to give it back
861 if (p->p_oppid != p->p_pptr->p_pid) {
863 t = proc_realparent(p);
867 "wait: traced child %d moved back to parent %d", p->p_pid,
869 proc_reparent(p, t, false);
871 pksignal(t, SIGCHLD, p->p_ksi);
873 cv_broadcast(&p->p_pwait);
875 sx_xunlock(&proctree_lock);
881 * Remove other references to this process to ensure we have an
882 * exclusive reference.
884 sx_xlock(&allproc_lock);
885 LIST_REMOVE(p, p_list); /* off zombproc */
886 sx_xunlock(&allproc_lock);
887 LIST_REMOVE(p, p_sibling);
888 reaper_abandon_children(p, true);
889 LIST_REMOVE(p, p_reapsibling);
891 proc_clear_orphan(p);
894 if (p->p_procdesc != NULL)
896 sx_xunlock(&proctree_lock);
899 knlist_detach(p->p_klist);
904 * Removal from allproc list and process group list paired with
905 * PROC_LOCK which was executed during that time should guarantee
906 * nothing can reach this process anymore. As such further locking
909 p->p_xexit = p->p_xsig = 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.
936 proc_set_cred(p, NULL);
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.
956 KASSERT(FIRST_THREAD_IN_PROC(p),
957 ("proc_reap: no residual thread!"));
958 uma_zfree(proc_zone, p);
959 atomic_add_int(&nprocs, -1);
963 proc_to_reap(struct thread *td, struct proc *p, idtype_t idtype, id_t id,
964 int *status, int options, struct __wrusage *wrusage, siginfo_t *siginfo,
969 sx_assert(&proctree_lock, SA_XLOCKED);
975 if (p->p_procdesc != NULL) {
981 if (p->p_pid != (pid_t)id) {
987 if (p->p_pgid != (pid_t)id) {
993 if (p->p_session->s_sid != (pid_t)id) {
999 if (p->p_ucred->cr_uid != (uid_t)id) {
1005 if (p->p_ucred->cr_gid != (gid_t)id) {
1011 if (p->p_ucred->cr_prison->pr_id != (int)id) {
1017 * It seems that the thread structures get zeroed out
1018 * at process exit. This makes it impossible to
1019 * support P_SETID, P_CID or P_CPUID.
1026 if (p_canwait(td, p)) {
1031 if (((options & WEXITED) == 0) && (p->p_state == PRS_ZOMBIE)) {
1037 * This special case handles a kthread spawned by linux_clone
1038 * (see linux_misc.c). The linux_wait4 and linux_waitpid
1039 * functions need to be able to distinguish between waiting
1040 * on a process and waiting on a thread. It is a thread if
1041 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option
1042 * signifies we want to wait for threads and not processes.
1044 if ((p->p_sigparent != SIGCHLD) ^
1045 ((options & WLINUXCLONE) != 0)) {
1050 if (siginfo != NULL) {
1051 bzero(siginfo, sizeof(*siginfo));
1052 siginfo->si_errno = 0;
1055 * SUSv4 requires that the si_signo value is always
1056 * SIGCHLD. Obey it despite the rfork(2) interface
1057 * allows to request other signal for child exit
1060 siginfo->si_signo = SIGCHLD;
1063 * This is still a rough estimate. We will fix the
1064 * cases TRAPPED, STOPPED, and CONTINUED later.
1066 if (WCOREDUMP(p->p_xsig)) {
1067 siginfo->si_code = CLD_DUMPED;
1068 siginfo->si_status = WTERMSIG(p->p_xsig);
1069 } else if (WIFSIGNALED(p->p_xsig)) {
1070 siginfo->si_code = CLD_KILLED;
1071 siginfo->si_status = WTERMSIG(p->p_xsig);
1073 siginfo->si_code = CLD_EXITED;
1074 siginfo->si_status = p->p_xexit;
1077 siginfo->si_pid = p->p_pid;
1078 siginfo->si_uid = p->p_ucred->cr_uid;
1081 * The si_addr field would be useful additional
1082 * detail, but apparently the PC value may be lost
1083 * when we reach this point. bzero() above sets
1084 * siginfo->si_addr to NULL.
1089 * There should be no reason to limit resources usage info to
1090 * exited processes only. A snapshot about any resources used
1091 * by a stopped process may be exactly what is needed.
1093 if (wrusage != NULL) {
1094 rup = &wrusage->wru_self;
1097 calcru(p, &rup->ru_utime, &rup->ru_stime);
1100 rup = &wrusage->wru_children;
1101 *rup = p->p_stats->p_cru;
1102 calccru(p, &rup->ru_utime, &rup->ru_stime);
1105 if (p->p_state == PRS_ZOMBIE && !check_only) {
1106 proc_reap(td, p, status, options);
1113 kern_wait(struct thread *td, pid_t pid, int *status, int options,
1114 struct rusage *rusage)
1116 struct __wrusage wru, *wrup;
1122 * Translate the special pid values into the (idtype, pid)
1123 * pair for kern_wait6. The WAIT_MYPGRP case is handled by
1124 * kern_wait6() on its own.
1126 if (pid == WAIT_ANY) {
1129 } else if (pid < 0) {
1143 * For backward compatibility we implicitly add flags WEXITED
1144 * and WTRAPPED here.
1146 options |= WEXITED | WTRAPPED;
1147 ret = kern_wait6(td, idtype, id, status, options, wrup, NULL);
1149 *rusage = wru.wru_self;
1154 report_alive_proc(struct thread *td, struct proc *p, siginfo_t *siginfo,
1155 int *status, int options, int si_code)
1159 PROC_LOCK_ASSERT(p, MA_OWNED);
1160 sx_assert(&proctree_lock, SA_XLOCKED);
1161 MPASS(si_code == CLD_TRAPPED || si_code == CLD_STOPPED ||
1162 si_code == CLD_CONTINUED);
1164 cont = si_code == CLD_CONTINUED;
1165 if ((options & WNOWAIT) == 0) {
1167 p->p_flag &= ~P_CONTINUED;
1169 p->p_flag |= P_WAITED;
1170 PROC_LOCK(td->td_proc);
1171 sigqueue_take(p->p_ksi);
1172 PROC_UNLOCK(td->td_proc);
1174 sx_xunlock(&proctree_lock);
1175 if (siginfo != NULL) {
1176 siginfo->si_code = si_code;
1177 siginfo->si_status = cont ? SIGCONT : p->p_xsig;
1180 *status = cont ? SIGCONT : W_STOPCODE(p->p_xsig);
1182 td->td_retval[0] = p->p_pid;
1186 kern_wait6(struct thread *td, idtype_t idtype, id_t id, int *status,
1187 int options, struct __wrusage *wrusage, siginfo_t *siginfo)
1191 int error, nfound, ret;
1194 AUDIT_ARG_VALUE((int)idtype); /* XXX - This is likely wrong! */
1195 AUDIT_ARG_PID((pid_t)id); /* XXX - This may be wrong! */
1196 AUDIT_ARG_VALUE(options);
1200 if ((pid_t)id == WAIT_MYPGRP && (idtype == P_PID || idtype == P_PGID)) {
1202 id = (id_t)q->p_pgid;
1207 /* If we don't know the option, just return. */
1208 if ((options & ~(WUNTRACED | WNOHANG | WCONTINUED | WNOWAIT |
1209 WEXITED | WTRAPPED | WLINUXCLONE)) != 0)
1211 if ((options & (WEXITED | WUNTRACED | WCONTINUED | WTRAPPED)) == 0) {
1213 * We will be unable to find any matching processes,
1214 * because there are no known events to look for.
1215 * Prefer to return error instead of blocking
1222 if (q->p_flag & P_STATCHILD) {
1224 q->p_flag &= ~P_STATCHILD;
1227 sx_xlock(&proctree_lock);
1230 LIST_FOREACH(p, &q->p_children, p_sibling) {
1232 ret = proc_to_reap(td, p, idtype, id, status, options,
1233 wrusage, siginfo, 0);
1236 else if (ret != 1) {
1237 td->td_retval[0] = pid;
1242 PROC_LOCK_ASSERT(p, MA_OWNED);
1244 if ((options & WTRAPPED) != 0 &&
1245 (p->p_flag & P_TRACED) != 0) {
1248 ((p->p_flag & (P_STOPPED_TRACE | P_STOPPED_SIG)) &&
1249 p->p_suspcount == p->p_numthreads &&
1250 (p->p_flag & P_WAITED) == 0);
1254 "wait: returning trapped pid %d status %#x "
1255 "(xstat %d) xthread %d",
1256 p->p_pid, W_STOPCODE(p->p_xsig), p->p_xsig,
1257 p->p_xthread != NULL ?
1258 p->p_xthread->td_tid : -1);
1259 report_alive_proc(td, p, siginfo, status,
1260 options, CLD_TRAPPED);
1264 if ((options & WUNTRACED) != 0 &&
1265 (p->p_flag & P_STOPPED_SIG) != 0) {
1267 report = (p->p_suspcount == p->p_numthreads &&
1268 ((p->p_flag & P_WAITED) == 0));
1271 report_alive_proc(td, p, siginfo, status,
1272 options, CLD_STOPPED);
1276 if ((options & WCONTINUED) != 0 &&
1277 (p->p_flag & P_CONTINUED) != 0) {
1278 report_alive_proc(td, p, siginfo, status, options,
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.
1298 LIST_FOREACH(p, &q->p_orphans, p_orphan) {
1299 ret = proc_to_reap(td, p, idtype, id, NULL, options,
1302 KASSERT(ret != -1, ("reaped an orphan (pid %d)",
1303 (int)td->td_retval[0]));
1311 sx_xunlock(&proctree_lock);
1314 if (options & WNOHANG) {
1315 sx_xunlock(&proctree_lock);
1316 td->td_retval[0] = 0;
1320 if (q->p_flag & P_STATCHILD) {
1321 q->p_flag &= ~P_STATCHILD;
1325 sx_xunlock(&proctree_lock);
1326 error = msleep(q, &q->p_mtx, PWAIT | PCATCH | PDROP, "wait", 0);
1333 proc_add_orphan(struct proc *child, struct proc *parent)
1336 sx_assert(&proctree_lock, SX_XLOCKED);
1337 KASSERT((child->p_flag & P_TRACED) != 0,
1338 ("proc_add_orphan: not traced"));
1340 if (LIST_EMPTY(&parent->p_orphans)) {
1341 child->p_treeflag |= P_TREE_FIRST_ORPHAN;
1342 LIST_INSERT_HEAD(&parent->p_orphans, child, p_orphan);
1344 LIST_INSERT_AFTER(LIST_FIRST(&parent->p_orphans),
1347 child->p_treeflag |= P_TREE_ORPHANED;
1351 * Make process 'parent' the new parent of process 'child'.
1352 * Must be called with an exclusive hold of proctree lock.
1355 proc_reparent(struct proc *child, struct proc *parent, bool set_oppid)
1358 sx_assert(&proctree_lock, SX_XLOCKED);
1359 PROC_LOCK_ASSERT(child, MA_OWNED);
1360 if (child->p_pptr == parent)
1363 PROC_LOCK(child->p_pptr);
1364 sigqueue_take(child->p_ksi);
1365 PROC_UNLOCK(child->p_pptr);
1366 LIST_REMOVE(child, p_sibling);
1367 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
1369 proc_clear_orphan(child);
1370 if ((child->p_flag & P_TRACED) != 0) {
1371 proc_add_orphan(child, child->p_pptr);
1374 child->p_pptr = parent;
1376 child->p_oppid = parent->p_pid;