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>
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, , , 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 return (child->p_pptr->p_pid == child->p_oppid ?
111 child->p_pptr : initproc);
112 for (p = child; (p->p_treeflag & P_TREE_FIRST_ORPHAN) == 0;) {
113 /* Cannot use LIST_PREV(), since the list head is not known. */
114 p = __containerof(p->p_orphan.le_prev, struct proc,
116 KASSERT((p->p_treeflag & P_TREE_ORPHANED) != 0,
117 ("missing P_ORPHAN %p", p));
119 parent = __containerof(p->p_orphan.le_prev, struct proc,
125 reaper_abandon_children(struct proc *p, bool exiting)
127 struct proc *p1, *p2, *ptmp;
129 sx_assert(&proctree_lock, SX_LOCKED);
130 KASSERT(p != initproc, ("reaper_abandon_children for initproc"));
131 if ((p->p_treeflag & P_TREE_REAPER) == 0)
134 LIST_FOREACH_SAFE(p2, &p->p_reaplist, p_reapsibling, ptmp) {
135 LIST_REMOVE(p2, p_reapsibling);
137 p2->p_reapsubtree = p->p_reapsubtree;
138 LIST_INSERT_HEAD(&p1->p_reaplist, p2, p_reapsibling);
139 if (exiting && p2->p_pptr == p) {
141 proc_reparent(p2, p1, true);
145 KASSERT(LIST_EMPTY(&p->p_reaplist), ("p_reaplist not empty"));
146 p->p_treeflag &= ~P_TREE_REAPER;
150 reaper_clear(struct proc *p)
155 sx_assert(&proctree_lock, SX_LOCKED);
156 LIST_REMOVE(p, p_reapsibling);
157 if (p->p_reapsubtree == 1)
160 LIST_FOREACH(p1, &p->p_reaper->p_reaplist, p_reapsibling) {
161 if (p1->p_reapsubtree == p->p_reapsubtree) {
167 proc_id_clear(PROC_ID_REAP, p->p_reapsubtree);
171 clear_orphan(struct proc *p)
175 sx_assert(&proctree_lock, SA_XLOCKED);
176 if ((p->p_treeflag & P_TREE_ORPHANED) == 0)
178 if ((p->p_treeflag & P_TREE_FIRST_ORPHAN) != 0) {
179 p1 = LIST_NEXT(p, p_orphan);
181 p1->p_treeflag |= P_TREE_FIRST_ORPHAN;
182 p->p_treeflag &= ~P_TREE_FIRST_ORPHAN;
184 LIST_REMOVE(p, p_orphan);
185 p->p_treeflag &= ~P_TREE_ORPHANED;
189 * exit -- death of process.
192 sys_sys_exit(struct thread *td, struct sys_exit_args *uap)
195 exit1(td, uap->rval, 0);
200 * Exit: deallocate address space and other resources, change proc state to
201 * zombie, and unlink proc from allproc and parent's lists. Save exit status
202 * and rusage for wait(). Check for child processes and orphan them.
205 exit1(struct thread *td, int rval, int signo)
207 struct proc *p, *nq, *q, *t;
209 ksiginfo_t *ksi, *ksi1;
212 mtx_assert(&Giant, MA_NOTOWNED);
213 KASSERT(rval == 0 || signo == 0, ("exit1 rv %d sig %d", rval, signo));
217 * XXX in case we're rebooting we just let init die in order to
218 * work around an unsolved stack overflow seen very late during
219 * shutdown on sparc64 when the gmirror worker process exists.
221 if (p == initproc && rebooting == 0) {
222 printf("init died (signal %d, exit %d)\n", signo, rval);
223 panic("Going nowhere without my init!");
227 * Deref SU mp, since the thread does not return to userspace.
229 td_softdep_cleanup(td);
232 * MUST abort all other threads before proceeding past here.
236 * First check if some other thread or external request got
237 * here before us. If so, act appropriately: exit or suspend.
238 * We must ensure that stop requests are handled before we set
241 thread_suspend_check(0);
242 while (p->p_flag & P_HADTHREADS) {
244 * Kill off the other threads. This requires
245 * some co-operation from other parts of the kernel
246 * so it may not be instantaneous. With this state set
247 * any thread entering the kernel from userspace will
248 * thread_exit() in trap(). Any thread attempting to
249 * sleep will return immediately with EINTR or EWOULDBLOCK
250 * which will hopefully force them to back out to userland
251 * freeing resources as they go. Any thread attempting
252 * to return to userland will thread_exit() from userret().
253 * thread_exit() will unsuspend us when the last of the
254 * other threads exits.
255 * If there is already a thread singler after resumption,
256 * calling thread_single will fail; in that case, we just
257 * re-check all suspension request, the thread should
258 * either be suspended there or exit.
260 if (!thread_single(p, SINGLE_EXIT))
262 * All other activity in this process is now
263 * stopped. Threading support has been turned
268 * Recheck for new stop or suspend requests which
269 * might appear while process lock was dropped in
272 thread_suspend_check(0);
274 KASSERT(p->p_numthreads == 1,
275 ("exit1: proc %p exiting with %d threads", p, p->p_numthreads));
276 racct_sub(p, RACCT_NTHR, 1);
278 /* Let event handler change exit status */
283 * Wakeup anyone in procfs' PIOCWAIT. They should have a hold
284 * on our vmspace, so we should block below until they have
285 * released their reference to us. Note that if they have
286 * requested S_EXIT stops we will block here until they ack
289 _STOPEVENT(p, S_EXIT, 0);
292 * Ignore any pending request to stop due to a stop signal.
293 * Once P_WEXIT is set, future requests will be ignored as
296 p->p_flag &= ~P_STOPPED_SIG;
297 KASSERT(!P_SHOULDSTOP(p), ("exiting process is stopped"));
300 * Note that we are exiting and do another wakeup of anyone in
301 * PIOCWAIT in case they aren't listening for S_EXIT stops or
302 * decided to wait again after we told them we are exiting.
304 p->p_flag |= P_WEXIT;
308 * Wait for any processes that have a hold on our vmspace to
309 * release their reference.
311 while (p->p_lock > 0)
312 msleep(&p->p_lock, &p->p_mtx, PWAIT, "exithold", 0);
315 /* Drain the limit callout while we don't have the proc locked */
316 callout_drain(&p->p_limco);
320 * The Sun BSM exit token contains two components: an exit status as
321 * passed to exit(), and a return value to indicate what sort of exit
322 * it was. The exit status is WEXITSTATUS(rv), but it's not clear
323 * what the return value is.
325 AUDIT_ARG_EXIT(rval, 0);
326 AUDIT_SYSCALL_EXIT(0, td);
329 /* Are we a task leader with peers? */
330 if (p->p_peers != NULL && p == p->p_leader) {
331 mtx_lock(&ppeers_lock);
335 kern_psignal(q, SIGKILL);
339 while (p->p_peers != NULL)
340 msleep(p, &ppeers_lock, PWAIT, "exit1", 0);
341 mtx_unlock(&ppeers_lock);
345 * Check if any loadable modules need anything done at process exit.
346 * E.g. SYSV IPC stuff.
347 * Event handler could change exit status.
348 * XXX what if one of these generates an error?
350 EVENTHANDLER_DIRECT_INVOKE(process_exit, p);
353 * If parent is waiting for us to exit or exec,
354 * P_PPWAIT is set; we will wakeup the parent below.
358 p->p_flag &= ~(P_TRACED | P_PPWAIT | P_PPTRACE);
362 * Stop the real interval timer. If the handler is currently
363 * executing, prevent it from rearming itself and let it finish.
365 if (timevalisset(&p->p_realtimer.it_value) &&
366 _callout_stop_safe(&p->p_itcallout, CS_EXECUTING, NULL) == 0) {
367 timevalclear(&p->p_realtimer.it_interval);
368 msleep(&p->p_itcallout, &p->p_mtx, PWAIT, "ritwait", 0);
369 KASSERT(!timevalisset(&p->p_realtimer.it_value),
370 ("realtime timer is still armed"));
375 umtx_thread_exit(td);
378 * Reset any sigio structures pointing to us as a result of
379 * F_SETOWN with our pid.
381 funsetownlst(&p->p_sigiolst);
384 * If this process has an nlminfo data area (for lockd), release it
386 if (nlminfo_release_p != NULL && p->p_nlminfo != NULL)
387 (*nlminfo_release_p)(p);
390 * Close open files and release open-file table.
396 * If this thread tickled GEOM, we need to wait for the giggling to
397 * stop before we return to userland
399 if (td->td_pflags & TDP_GEOM)
403 * Remove ourself from our leader's peer list and wake our leader.
405 if (p->p_leader->p_peers != NULL) {
406 mtx_lock(&ppeers_lock);
407 if (p->p_leader->p_peers != NULL) {
409 while (q->p_peers != p)
411 q->p_peers = p->p_peers;
414 mtx_unlock(&ppeers_lock);
419 (void)acct_process(td);
425 * Release reference to text vnode
427 if (p->p_textvp != NULL) {
433 * Release our limits structure.
435 lim_free(p->p_limit);
441 * Call machine-dependent code to release any
442 * machine-dependent resources other than the address space.
443 * The address space is released by "vmspace_exitfree(p)" in
448 WITNESS_WARN(WARN_PANIC, NULL, "process (pid %d) exiting", p->p_pid);
451 * Move proc from allproc queue to zombproc.
453 sx_xlock(&allproc_lock);
454 sx_xlock(&zombproc_lock);
455 LIST_REMOVE(p, p_list);
456 LIST_INSERT_HEAD(&zombproc, p, p_list);
457 sx_xunlock(&zombproc_lock);
458 sx_xunlock(&allproc_lock);
460 sx_xlock(&proctree_lock);
463 * Reparent all children processes:
464 * - traced ones to the original parent (or init if we are that parent)
467 q = LIST_FIRST(&p->p_children);
468 if (q != NULL) /* only need this if any child is S_ZOMB */
470 for (; q != NULL; q = nq) {
471 nq = LIST_NEXT(q, p_sibling);
472 ksi = ksiginfo_alloc(TRUE);
474 q->p_sigparent = SIGCHLD;
476 if (!(q->p_flag & P_TRACED)) {
477 proc_reparent(q, q->p_reaper, true);
478 if (q->p_state == PRS_ZOMBIE) {
480 * Inform reaper about the reparented
481 * zombie, since wait(2) has something
482 * new to report. Guarantee queueing
483 * of the SIGCHLD signal, similar to
484 * the _exit() behaviour, by providing
485 * our ksiginfo. Ksi is freed by the
488 if (q->p_ksi == NULL) {
491 ksiginfo_copy(q->p_ksi, ksi);
492 ksi->ksi_flags |= KSI_INS;
496 PROC_LOCK(q->p_reaper);
497 pksignal(q->p_reaper, SIGCHLD, ksi1);
498 PROC_UNLOCK(q->p_reaper);
499 } else if (q->p_pdeathsig > 0) {
501 * The child asked to received a signal
504 kern_psignal(q, q->p_pdeathsig);
508 * Traced processes are killed since their existence
509 * means someone is screwing up.
511 t = proc_realparent(q);
513 proc_reparent(q, q->p_reaper, true);
516 proc_reparent(q, t, true);
520 * Since q was found on our children list, the
521 * proc_reparent() call moved q to the orphan
522 * list due to present P_TRACED flag. Clear
523 * orphan link for q now while q is locked.
526 q->p_flag &= ~(P_TRACED | P_STOPPED_TRACE);
527 q->p_flag2 &= ~P2_PTRACE_FSTP;
529 FOREACH_THREAD_IN_PROC(q, tdt) {
530 tdt->td_dbgflags &= ~(TDB_SUSPEND | TDB_XSIG |
533 kern_psignal(q, SIGKILL);
541 * Also get rid of our orphans.
543 while ((q = LIST_FIRST(&p->p_orphans)) != NULL) {
545 KASSERT(q->p_oppid == p->p_pid,
546 ("orphan %p of %p has unexpected oppid %d", q, p,
548 q->p_oppid = q->p_reaper->p_pid;
551 * If we are the real parent of this process
552 * but it has been reparented to a debugger, then
553 * check if it asked for a signal when we exit.
555 if (q->p_pdeathsig > 0)
556 kern_psignal(q, q->p_pdeathsig);
557 CTR2(KTR_PTRACE, "exit: pid %d, clearing orphan %d", p->p_pid,
564 if (SDT_PROBES_ENABLED()) {
565 int reason = CLD_EXITED;
566 if (WCOREDUMP(signo))
568 else if (WIFSIGNALED(signo))
570 SDT_PROBE1(proc, , , exit, reason);
574 /* Save exit status. */
580 * Tell the DTrace fasttrap provider about the exit if it
581 * has declared an interest.
583 if (dtrace_fasttrap_exit)
584 dtrace_fasttrap_exit(p);
588 * Notify interested parties of our demise.
590 KNOTE_LOCKED(p->p_klist, NOTE_EXIT);
593 * If this is a process with a descriptor, we may not need to deliver
594 * a signal to the parent. proctree_lock is held over
595 * procdesc_exit() to serialize concurrent calls to close() and
599 if (p->p_procdesc == NULL || procdesc_exit(p)) {
601 * Notify parent that we're gone. If parent has the
602 * PS_NOCLDWAIT flag set, or if the handler is set to SIG_IGN,
603 * notify process 1 instead (and hope it will handle this
606 PROC_LOCK(p->p_pptr);
607 mtx_lock(&p->p_pptr->p_sigacts->ps_mtx);
608 if (p->p_pptr->p_sigacts->ps_flag &
609 (PS_NOCLDWAIT | PS_CLDSIGIGN)) {
612 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
615 proc_reparent(p, p->p_reaper, true);
616 p->p_sigparent = SIGCHLD;
617 PROC_LOCK(p->p_pptr);
620 * Notify parent, so in case he was wait(2)ing or
621 * executing waitpid(2) with our pid, he will
626 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
628 if (p->p_pptr == p->p_reaper || p->p_pptr == initproc) {
630 } else if (p->p_sigparent != 0) {
631 if (p->p_sigparent == SIGCHLD) {
633 } else { /* LINUX thread */
638 PROC_LOCK(p->p_pptr);
639 sx_xunlock(&proctree_lock);
641 if (signal_parent == 1) {
643 } else if (signal_parent == 2) {
644 kern_psignal(p->p_pptr, p->p_sigparent);
647 /* Tell the prison that we are gone. */
648 prison_proc_free(p->p_ucred->cr_prison);
651 * The state PRS_ZOMBIE prevents other proesses from sending
652 * signal to the process, to avoid memory leak, we free memory
653 * for signal queue at the time when the state is set.
655 sigqueue_flush(&p->p_sigqueue);
656 sigqueue_flush(&td->td_sigqueue);
659 * We have to wait until after acquiring all locks before
660 * changing p_state. We need to avoid all possible context
661 * switches (including ones from blocking on a mutex) while
662 * marked as a zombie. We also have to set the zombie state
663 * before we release the parent process' proc lock to avoid
664 * a lost wakeup. So, we first call wakeup, then we grab the
665 * sched lock, update the state, and release the parent process'
669 cv_broadcast(&p->p_pwait);
670 sched_exit(p->p_pptr, td);
672 p->p_state = PRS_ZOMBIE;
673 PROC_UNLOCK(p->p_pptr);
676 * Save our children's rusage information in our exit rusage.
679 ruadd(&p->p_ru, &p->p_rux, &p->p_stats->p_cru, &p->p_crux);
683 * Make sure the scheduler takes this thread out of its tables etc.
684 * This will also release this thread's reference to the ucred.
685 * Other thread parts to release include pcb bits and such.
691 #ifndef _SYS_SYSPROTO_H_
700 sys_abort2(struct thread *td, struct abort2_args *uap)
702 struct proc *p = td->td_proc;
708 * Do it right now so we can log either proper call of abort2(), or
709 * note, that invalid argument was passed. 512 is big enough to
710 * handle 16 arguments' descriptions with additional comments.
712 sb = sbuf_new(NULL, NULL, 512, SBUF_FIXEDLEN);
714 sbuf_printf(sb, "%s(pid %d uid %d) aborted: ",
715 p->p_comm, p->p_pid, td->td_ucred->cr_uid);
717 * Since we can't return from abort2(), send SIGKILL in cases, where
718 * abort2() was called improperly
721 /* Prevent from DoSes from user-space. */
722 if (uap->nargs < 0 || uap->nargs > 16)
724 if (uap->nargs > 0) {
725 if (uap->args == NULL)
727 error = copyin(uap->args, uargs, uap->nargs * sizeof(void *));
732 * Limit size of 'reason' string to 128. Will fit even when
733 * maximal number of arguments was chosen to be logged.
735 if (uap->why != NULL) {
736 error = sbuf_copyin(sb, uap->why, 128);
740 sbuf_printf(sb, "(null)");
742 if (uap->nargs > 0) {
743 sbuf_printf(sb, "(");
744 for (i = 0;i < uap->nargs; i++)
745 sbuf_printf(sb, "%s%p", i == 0 ? "" : ", ", uargs[i]);
746 sbuf_printf(sb, ")");
749 * Final stage: arguments were proper, string has been
750 * successfully copied from userspace, and copying pointers
751 * from user-space succeed.
755 if (sig == SIGKILL) {
757 sbuf_printf(sb, " (Reason text inaccessible)");
761 log(LOG_INFO, "%s", sbuf_data(sb));
770 * The dirty work is handled by kern_wait().
773 owait(struct thread *td, struct owait_args *uap __unused)
777 error = kern_wait(td, WAIT_ANY, &status, 0, NULL);
779 td->td_retval[1] = status;
782 #endif /* COMPAT_43 */
785 * The dirty work is handled by kern_wait().
788 sys_wait4(struct thread *td, struct wait4_args *uap)
790 struct rusage ru, *rup;
793 if (uap->rusage != NULL)
797 error = kern_wait(td, uap->pid, &status, uap->options, rup);
798 if (uap->status != NULL && error == 0 && td->td_retval[0] != 0)
799 error = copyout(&status, uap->status, sizeof(status));
800 if (uap->rusage != NULL && error == 0 && td->td_retval[0] != 0)
801 error = copyout(&ru, uap->rusage, sizeof(struct rusage));
806 sys_wait6(struct thread *td, struct wait6_args *uap)
808 struct __wrusage wru, *wrup;
814 idtype = uap->idtype;
817 if (uap->wrusage != NULL)
822 if (uap->info != NULL) {
824 bzero(sip, sizeof(*sip));
829 * We expect all callers of wait6() to know about WEXITED and
832 error = kern_wait6(td, idtype, id, &status, uap->options, wrup, sip);
834 if (uap->status != NULL && error == 0 && td->td_retval[0] != 0)
835 error = copyout(&status, uap->status, sizeof(status));
836 if (uap->wrusage != NULL && error == 0 && td->td_retval[0] != 0)
837 error = copyout(&wru, uap->wrusage, sizeof(wru));
838 if (uap->info != NULL && error == 0)
839 error = copyout(&si, uap->info, sizeof(si));
844 * Reap the remains of a zombie process and optionally return status and
845 * rusage. Asserts and will release both the proctree_lock and the process
846 * lock as part of its work.
849 proc_reap(struct thread *td, struct proc *p, int *status, int options)
853 sx_assert(&proctree_lock, SA_XLOCKED);
854 PROC_LOCK_ASSERT(p, MA_OWNED);
855 KASSERT(p->p_state == PRS_ZOMBIE, ("proc_reap: !PRS_ZOMBIE"));
857 mtx_spin_wait_unlocked(&p->p_slock);
862 *status = KW_EXITCODE(p->p_xexit, p->p_xsig);
863 if (options & WNOWAIT) {
865 * Only poll, returning the status. Caller does not wish to
866 * release the proc struct just yet.
869 sx_xunlock(&proctree_lock);
874 sigqueue_take(p->p_ksi);
878 * If we got the child via a ptrace 'attach', we need to give it back
881 if (p->p_oppid != p->p_pptr->p_pid) {
883 t = proc_realparent(p);
887 "wait: traced child %d moved back to parent %d", p->p_pid,
889 proc_reparent(p, t, false);
891 pksignal(t, SIGCHLD, p->p_ksi);
893 cv_broadcast(&p->p_pwait);
895 sx_xunlock(&proctree_lock);
901 * Remove other references to this process to ensure we have an
902 * exclusive reference.
904 sx_xlock(&zombproc_lock);
905 LIST_REMOVE(p, p_list); /* off zombproc */
906 sx_xunlock(&zombproc_lock);
907 sx_xlock(PIDHASHLOCK(p->p_pid));
908 LIST_REMOVE(p, p_hash);
909 sx_xunlock(PIDHASHLOCK(p->p_pid));
910 LIST_REMOVE(p, p_sibling);
911 reaper_abandon_children(p, true);
913 proc_id_clear(PROC_ID_PID, p->p_pid);
918 if (p->p_procdesc != NULL)
920 sx_xunlock(&proctree_lock);
923 knlist_detach(p->p_klist);
928 * Removal from allproc list and process group list paired with
929 * PROC_LOCK which was executed during that time should guarantee
930 * nothing can reach this process anymore. As such further locking
933 p->p_xexit = p->p_xsig = 0; /* XXX: why? */
936 ruadd(&q->p_stats->p_cru, &q->p_crux, &p->p_ru, &p->p_rux);
940 * Decrement the count of procs running with this uid.
942 (void)chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0);
945 * Destroy resource accounting information associated with the process.
950 racct_sub(p, RACCT_NPROC, 1);
957 * Free credentials, arguments, and sigacts.
960 proc_set_cred(p, NULL);
961 pargs_drop(p->p_args);
963 sigacts_free(p->p_sigacts);
967 * Do any thread-system specific cleanups.
972 * Give vm and machine-dependent layer a chance to free anything that
973 * cpu_exit couldn't release while still running in process context.
980 KASSERT(FIRST_THREAD_IN_PROC(p),
981 ("proc_reap: no residual thread!"));
982 uma_zfree(proc_zone, p);
983 atomic_add_int(&nprocs, -1);
987 proc_to_reap(struct thread *td, struct proc *p, idtype_t idtype, id_t id,
988 int *status, int options, struct __wrusage *wrusage, siginfo_t *siginfo,
993 sx_assert(&proctree_lock, SA_XLOCKED);
999 if (p->p_procdesc != NULL) {
1005 if (p->p_pid != (pid_t)id) {
1011 if (p->p_pgid != (pid_t)id) {
1017 if (p->p_session->s_sid != (pid_t)id) {
1023 if (p->p_ucred->cr_uid != (uid_t)id) {
1029 if (p->p_ucred->cr_gid != (gid_t)id) {
1035 if (p->p_ucred->cr_prison->pr_id != (int)id) {
1041 * It seems that the thread structures get zeroed out
1042 * at process exit. This makes it impossible to
1043 * support P_SETID, P_CID or P_CPUID.
1050 if (p_canwait(td, p)) {
1055 if (((options & WEXITED) == 0) && (p->p_state == PRS_ZOMBIE)) {
1061 * This special case handles a kthread spawned by linux_clone
1062 * (see linux_misc.c). The linux_wait4 and linux_waitpid
1063 * functions need to be able to distinguish between waiting
1064 * on a process and waiting on a thread. It is a thread if
1065 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option
1066 * signifies we want to wait for threads and not processes.
1068 if ((p->p_sigparent != SIGCHLD) ^
1069 ((options & WLINUXCLONE) != 0)) {
1074 if (siginfo != NULL) {
1075 bzero(siginfo, sizeof(*siginfo));
1076 siginfo->si_errno = 0;
1079 * SUSv4 requires that the si_signo value is always
1080 * SIGCHLD. Obey it despite the rfork(2) interface
1081 * allows to request other signal for child exit
1084 siginfo->si_signo = SIGCHLD;
1087 * This is still a rough estimate. We will fix the
1088 * cases TRAPPED, STOPPED, and CONTINUED later.
1090 if (WCOREDUMP(p->p_xsig)) {
1091 siginfo->si_code = CLD_DUMPED;
1092 siginfo->si_status = WTERMSIG(p->p_xsig);
1093 } else if (WIFSIGNALED(p->p_xsig)) {
1094 siginfo->si_code = CLD_KILLED;
1095 siginfo->si_status = WTERMSIG(p->p_xsig);
1097 siginfo->si_code = CLD_EXITED;
1098 siginfo->si_status = p->p_xexit;
1101 siginfo->si_pid = p->p_pid;
1102 siginfo->si_uid = p->p_ucred->cr_uid;
1105 * The si_addr field would be useful additional
1106 * detail, but apparently the PC value may be lost
1107 * when we reach this point. bzero() above sets
1108 * siginfo->si_addr to NULL.
1113 * There should be no reason to limit resources usage info to
1114 * exited processes only. A snapshot about any resources used
1115 * by a stopped process may be exactly what is needed.
1117 if (wrusage != NULL) {
1118 rup = &wrusage->wru_self;
1121 calcru(p, &rup->ru_utime, &rup->ru_stime);
1124 rup = &wrusage->wru_children;
1125 *rup = p->p_stats->p_cru;
1126 calccru(p, &rup->ru_utime, &rup->ru_stime);
1129 if (p->p_state == PRS_ZOMBIE && !check_only) {
1130 proc_reap(td, p, status, options);
1137 kern_wait(struct thread *td, pid_t pid, int *status, int options,
1138 struct rusage *rusage)
1140 struct __wrusage wru, *wrup;
1146 * Translate the special pid values into the (idtype, pid)
1147 * pair for kern_wait6. The WAIT_MYPGRP case is handled by
1148 * kern_wait6() on its own.
1150 if (pid == WAIT_ANY) {
1153 } else if (pid < 0) {
1167 * For backward compatibility we implicitly add flags WEXITED
1168 * and WTRAPPED here.
1170 options |= WEXITED | WTRAPPED;
1171 ret = kern_wait6(td, idtype, id, status, options, wrup, NULL);
1173 *rusage = wru.wru_self;
1178 report_alive_proc(struct thread *td, struct proc *p, siginfo_t *siginfo,
1179 int *status, int options, int si_code)
1183 PROC_LOCK_ASSERT(p, MA_OWNED);
1184 sx_assert(&proctree_lock, SA_XLOCKED);
1185 MPASS(si_code == CLD_TRAPPED || si_code == CLD_STOPPED ||
1186 si_code == CLD_CONTINUED);
1188 cont = si_code == CLD_CONTINUED;
1189 if ((options & WNOWAIT) == 0) {
1191 p->p_flag &= ~P_CONTINUED;
1193 p->p_flag |= P_WAITED;
1194 PROC_LOCK(td->td_proc);
1195 sigqueue_take(p->p_ksi);
1196 PROC_UNLOCK(td->td_proc);
1198 sx_xunlock(&proctree_lock);
1199 if (siginfo != NULL) {
1200 siginfo->si_code = si_code;
1201 siginfo->si_status = cont ? SIGCONT : p->p_xsig;
1204 *status = cont ? SIGCONT : W_STOPCODE(p->p_xsig);
1206 td->td_retval[0] = p->p_pid;
1210 kern_wait6(struct thread *td, idtype_t idtype, id_t id, int *status,
1211 int options, struct __wrusage *wrusage, siginfo_t *siginfo)
1215 int error, nfound, ret;
1218 AUDIT_ARG_VALUE((int)idtype); /* XXX - This is likely wrong! */
1219 AUDIT_ARG_PID((pid_t)id); /* XXX - This may be wrong! */
1220 AUDIT_ARG_VALUE(options);
1224 if ((pid_t)id == WAIT_MYPGRP && (idtype == P_PID || idtype == P_PGID)) {
1226 id = (id_t)q->p_pgid;
1231 /* If we don't know the option, just return. */
1232 if ((options & ~(WUNTRACED | WNOHANG | WCONTINUED | WNOWAIT |
1233 WEXITED | WTRAPPED | WLINUXCLONE)) != 0)
1235 if ((options & (WEXITED | WUNTRACED | WCONTINUED | WTRAPPED)) == 0) {
1237 * We will be unable to find any matching processes,
1238 * because there are no known events to look for.
1239 * Prefer to return error instead of blocking
1246 if (q->p_flag & P_STATCHILD) {
1248 q->p_flag &= ~P_STATCHILD;
1251 sx_xlock(&proctree_lock);
1254 LIST_FOREACH(p, &q->p_children, p_sibling) {
1256 ret = proc_to_reap(td, p, idtype, id, status, options,
1257 wrusage, siginfo, 0);
1260 else if (ret != 1) {
1261 td->td_retval[0] = pid;
1266 PROC_LOCK_ASSERT(p, MA_OWNED);
1268 if ((options & WTRAPPED) != 0 &&
1269 (p->p_flag & P_TRACED) != 0) {
1272 ((p->p_flag & (P_STOPPED_TRACE | P_STOPPED_SIG)) &&
1273 p->p_suspcount == p->p_numthreads &&
1274 (p->p_flag & P_WAITED) == 0);
1278 "wait: returning trapped pid %d status %#x "
1279 "(xstat %d) xthread %d",
1280 p->p_pid, W_STOPCODE(p->p_xsig), p->p_xsig,
1281 p->p_xthread != NULL ?
1282 p->p_xthread->td_tid : -1);
1283 report_alive_proc(td, p, siginfo, status,
1284 options, CLD_TRAPPED);
1288 if ((options & WUNTRACED) != 0 &&
1289 (p->p_flag & P_STOPPED_SIG) != 0) {
1291 report = (p->p_suspcount == p->p_numthreads &&
1292 ((p->p_flag & P_WAITED) == 0));
1295 report_alive_proc(td, p, siginfo, status,
1296 options, CLD_STOPPED);
1300 if ((options & WCONTINUED) != 0 &&
1301 (p->p_flag & P_CONTINUED) != 0) {
1302 report_alive_proc(td, p, siginfo, status, options,
1310 * Look in the orphans list too, to allow the parent to
1311 * collect it's child exit status even if child is being
1314 * Debugger detaches from the parent upon successful
1315 * switch-over from parent to child. At this point due to
1316 * re-parenting the parent loses the child to debugger and a
1317 * wait4(2) call would report that it has no children to wait
1318 * for. By maintaining a list of orphans we allow the parent
1319 * to successfully wait until the child becomes a zombie.
1322 LIST_FOREACH(p, &q->p_orphans, p_orphan) {
1323 ret = proc_to_reap(td, p, idtype, id, NULL, options,
1326 KASSERT(ret != -1, ("reaped an orphan (pid %d)",
1327 (int)td->td_retval[0]));
1335 sx_xunlock(&proctree_lock);
1338 if (options & WNOHANG) {
1339 sx_xunlock(&proctree_lock);
1340 td->td_retval[0] = 0;
1344 if (q->p_flag & P_STATCHILD) {
1345 q->p_flag &= ~P_STATCHILD;
1349 sx_xunlock(&proctree_lock);
1350 error = msleep(q, &q->p_mtx, PWAIT | PCATCH | PDROP, "wait", 0);
1357 * Make process 'parent' the new parent of process 'child'.
1358 * Must be called with an exclusive hold of proctree lock.
1361 proc_reparent(struct proc *child, struct proc *parent, bool set_oppid)
1364 sx_assert(&proctree_lock, SX_XLOCKED);
1365 PROC_LOCK_ASSERT(child, MA_OWNED);
1366 if (child->p_pptr == parent)
1369 PROC_LOCK(child->p_pptr);
1370 sigqueue_take(child->p_ksi);
1371 PROC_UNLOCK(child->p_pptr);
1372 LIST_REMOVE(child, p_sibling);
1373 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
1375 clear_orphan(child);
1376 if (child->p_flag & P_TRACED) {
1377 if (LIST_EMPTY(&child->p_pptr->p_orphans)) {
1378 child->p_treeflag |= P_TREE_FIRST_ORPHAN;
1379 LIST_INSERT_HEAD(&child->p_pptr->p_orphans, child,
1382 LIST_INSERT_AFTER(LIST_FIRST(&child->p_pptr->p_orphans),
1385 child->p_treeflag |= P_TREE_ORPHANED;
1388 child->p_pptr = parent;
1390 child->p_oppid = parent->p_pid;