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 proc_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.
361 * Stop the real interval timer. If the handler is currently
362 * executing, prevent it from rearming itself and let it finish.
364 if (timevalisset(&p->p_realtimer.it_value) &&
365 _callout_stop_safe(&p->p_itcallout, CS_EXECUTING, NULL) == 0) {
366 timevalclear(&p->p_realtimer.it_interval);
367 msleep(&p->p_itcallout, &p->p_mtx, PWAIT, "ritwait", 0);
368 KASSERT(!timevalisset(&p->p_realtimer.it_value),
369 ("realtime timer is still armed"));
374 umtx_thread_exit(td);
377 * Reset any sigio structures pointing to us as a result of
378 * F_SETOWN with our pid.
380 funsetownlst(&p->p_sigiolst);
383 * If this process has an nlminfo data area (for lockd), release it
385 if (nlminfo_release_p != NULL && p->p_nlminfo != NULL)
386 (*nlminfo_release_p)(p);
389 * Close open files and release open-file table.
395 * If this thread tickled GEOM, we need to wait for the giggling to
396 * stop before we return to userland
398 if (td->td_pflags & TDP_GEOM)
402 * Remove ourself from our leader's peer list and wake our leader.
404 if (p->p_leader->p_peers != NULL) {
405 mtx_lock(&ppeers_lock);
406 if (p->p_leader->p_peers != NULL) {
408 while (q->p_peers != p)
410 q->p_peers = p->p_peers;
413 mtx_unlock(&ppeers_lock);
418 (void)acct_process(td);
424 * Release reference to text vnode
426 if (p->p_textvp != NULL) {
432 * Release our limits structure.
434 lim_free(p->p_limit);
440 * Call machine-dependent code to release any
441 * machine-dependent resources other than the address space.
442 * The address space is released by "vmspace_exitfree(p)" in
447 WITNESS_WARN(WARN_PANIC, NULL, "process (pid %d) exiting", p->p_pid);
450 * Move proc from allproc queue to zombproc.
452 sx_xlock(&allproc_lock);
453 sx_xlock(&zombproc_lock);
454 LIST_REMOVE(p, p_list);
455 LIST_INSERT_HEAD(&zombproc, p, p_list);
456 sx_xunlock(&zombproc_lock);
457 sx_xunlock(&allproc_lock);
459 sx_xlock(&proctree_lock);
461 p->p_flag &= ~(P_TRACED | P_PPWAIT | P_PPTRACE);
465 * Reparent all children processes:
466 * - traced ones to the original parent (or init if we are that parent)
469 q = LIST_FIRST(&p->p_children);
470 if (q != NULL) /* only need this if any child is S_ZOMB */
472 for (; q != NULL; q = nq) {
473 nq = LIST_NEXT(q, p_sibling);
474 ksi = ksiginfo_alloc(TRUE);
476 q->p_sigparent = SIGCHLD;
478 if ((q->p_flag & P_TRACED) == 0) {
479 proc_reparent(q, q->p_reaper, true);
480 if (q->p_state == PRS_ZOMBIE) {
482 * Inform reaper about the reparented
483 * zombie, since wait(2) has something
484 * new to report. Guarantee queueing
485 * of the SIGCHLD signal, similar to
486 * the _exit() behaviour, by providing
487 * our ksiginfo. Ksi is freed by the
490 if (q->p_ksi == NULL) {
493 ksiginfo_copy(q->p_ksi, ksi);
494 ksi->ksi_flags |= KSI_INS;
498 PROC_LOCK(q->p_reaper);
499 pksignal(q->p_reaper, SIGCHLD, ksi1);
500 PROC_UNLOCK(q->p_reaper);
501 } else if (q->p_pdeathsig > 0) {
503 * The child asked to received a signal
506 kern_psignal(q, q->p_pdeathsig);
510 * Traced processes are killed since their existence
511 * means someone is screwing up.
513 t = proc_realparent(q);
515 proc_reparent(q, q->p_reaper, true);
518 proc_reparent(q, t, true);
522 * Since q was found on our children list, the
523 * proc_reparent() call moved q to the orphan
524 * list due to present P_TRACED flag. Clear
525 * orphan link for q now while q is locked.
527 proc_clear_orphan(q);
528 q->p_flag &= ~(P_TRACED | P_STOPPED_TRACE);
529 q->p_flag2 &= ~P2_PTRACE_FSTP;
531 FOREACH_THREAD_IN_PROC(q, tdt) {
532 tdt->td_dbgflags &= ~(TDB_SUSPEND | TDB_XSIG |
535 kern_psignal(q, SIGKILL);
543 * Also get rid of our orphans.
545 while ((q = LIST_FIRST(&p->p_orphans)) != NULL) {
547 KASSERT(q->p_oppid == p->p_pid,
548 ("orphan %p of %p has unexpected oppid %d", q, p,
550 q->p_oppid = q->p_reaper->p_pid;
553 * If we are the real parent of this process
554 * but it has been reparented to a debugger, then
555 * check if it asked for a signal when we exit.
557 if (q->p_pdeathsig > 0)
558 kern_psignal(q, q->p_pdeathsig);
559 CTR2(KTR_PTRACE, "exit: pid %d, clearing orphan %d", p->p_pid,
561 proc_clear_orphan(q);
566 if (SDT_PROBES_ENABLED()) {
567 int reason = CLD_EXITED;
568 if (WCOREDUMP(signo))
570 else if (WIFSIGNALED(signo))
572 SDT_PROBE1(proc, , , exit, reason);
576 /* Save exit status. */
582 * Tell the DTrace fasttrap provider about the exit if it
583 * has declared an interest.
585 if (dtrace_fasttrap_exit)
586 dtrace_fasttrap_exit(p);
590 * Notify interested parties of our demise.
592 KNOTE_LOCKED(p->p_klist, NOTE_EXIT);
595 * If this is a process with a descriptor, we may not need to deliver
596 * a signal to the parent. proctree_lock is held over
597 * procdesc_exit() to serialize concurrent calls to close() and
601 if (p->p_procdesc == NULL || procdesc_exit(p)) {
603 * Notify parent that we're gone. If parent has the
604 * PS_NOCLDWAIT flag set, or if the handler is set to SIG_IGN,
605 * notify process 1 instead (and hope it will handle this
608 PROC_LOCK(p->p_pptr);
609 mtx_lock(&p->p_pptr->p_sigacts->ps_mtx);
610 if (p->p_pptr->p_sigacts->ps_flag &
611 (PS_NOCLDWAIT | PS_CLDSIGIGN)) {
614 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
617 proc_reparent(p, p->p_reaper, true);
618 p->p_sigparent = SIGCHLD;
619 PROC_LOCK(p->p_pptr);
622 * Notify parent, so in case he was wait(2)ing or
623 * executing waitpid(2) with our pid, he will
628 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
630 if (p->p_pptr == p->p_reaper || p->p_pptr == initproc) {
632 } else if (p->p_sigparent != 0) {
633 if (p->p_sigparent == SIGCHLD) {
635 } else { /* LINUX thread */
640 PROC_LOCK(p->p_pptr);
641 sx_xunlock(&proctree_lock);
643 if (signal_parent == 1) {
645 } else if (signal_parent == 2) {
646 kern_psignal(p->p_pptr, p->p_sigparent);
649 /* Tell the prison that we are gone. */
650 prison_proc_free(p->p_ucred->cr_prison);
653 * The state PRS_ZOMBIE prevents other proesses from sending
654 * signal to the process, to avoid memory leak, we free memory
655 * for signal queue at the time when the state is set.
657 sigqueue_flush(&p->p_sigqueue);
658 sigqueue_flush(&td->td_sigqueue);
661 * We have to wait until after acquiring all locks before
662 * changing p_state. We need to avoid all possible context
663 * switches (including ones from blocking on a mutex) while
664 * marked as a zombie. We also have to set the zombie state
665 * before we release the parent process' proc lock to avoid
666 * a lost wakeup. So, we first call wakeup, then we grab the
667 * sched lock, update the state, and release the parent process'
671 cv_broadcast(&p->p_pwait);
672 sched_exit(p->p_pptr, td);
674 p->p_state = PRS_ZOMBIE;
675 PROC_UNLOCK(p->p_pptr);
678 * Save our children's rusage information in our exit rusage.
681 ruadd(&p->p_ru, &p->p_rux, &p->p_stats->p_cru, &p->p_crux);
685 * Make sure the scheduler takes this thread out of its tables etc.
686 * This will also release this thread's reference to the ucred.
687 * Other thread parts to release include pcb bits and such.
693 #ifndef _SYS_SYSPROTO_H_
702 sys_abort2(struct thread *td, struct abort2_args *uap)
704 struct proc *p = td->td_proc;
710 * Do it right now so we can log either proper call of abort2(), or
711 * note, that invalid argument was passed. 512 is big enough to
712 * handle 16 arguments' descriptions with additional comments.
714 sb = sbuf_new(NULL, NULL, 512, SBUF_FIXEDLEN);
716 sbuf_printf(sb, "%s(pid %d uid %d) aborted: ",
717 p->p_comm, p->p_pid, td->td_ucred->cr_uid);
719 * Since we can't return from abort2(), send SIGKILL in cases, where
720 * abort2() was called improperly
723 /* Prevent from DoSes from user-space. */
724 if (uap->nargs < 0 || uap->nargs > 16)
726 if (uap->nargs > 0) {
727 if (uap->args == NULL)
729 error = copyin(uap->args, uargs, uap->nargs * sizeof(void *));
734 * Limit size of 'reason' string to 128. Will fit even when
735 * maximal number of arguments was chosen to be logged.
737 if (uap->why != NULL) {
738 error = sbuf_copyin(sb, uap->why, 128);
742 sbuf_printf(sb, "(null)");
744 if (uap->nargs > 0) {
745 sbuf_printf(sb, "(");
746 for (i = 0;i < uap->nargs; i++)
747 sbuf_printf(sb, "%s%p", i == 0 ? "" : ", ", uargs[i]);
748 sbuf_printf(sb, ")");
751 * Final stage: arguments were proper, string has been
752 * successfully copied from userspace, and copying pointers
753 * from user-space succeed.
757 if (sig == SIGKILL) {
759 sbuf_printf(sb, " (Reason text inaccessible)");
763 log(LOG_INFO, "%s", sbuf_data(sb));
772 * The dirty work is handled by kern_wait().
775 owait(struct thread *td, struct owait_args *uap __unused)
779 error = kern_wait(td, WAIT_ANY, &status, 0, NULL);
781 td->td_retval[1] = status;
784 #endif /* COMPAT_43 */
787 * The dirty work is handled by kern_wait().
790 sys_wait4(struct thread *td, struct wait4_args *uap)
792 struct rusage ru, *rup;
795 if (uap->rusage != NULL)
799 error = kern_wait(td, uap->pid, &status, uap->options, rup);
800 if (uap->status != NULL && error == 0 && td->td_retval[0] != 0)
801 error = copyout(&status, uap->status, sizeof(status));
802 if (uap->rusage != NULL && error == 0 && td->td_retval[0] != 0)
803 error = copyout(&ru, uap->rusage, sizeof(struct rusage));
808 sys_wait6(struct thread *td, struct wait6_args *uap)
810 struct __wrusage wru, *wrup;
816 idtype = uap->idtype;
819 if (uap->wrusage != NULL)
824 if (uap->info != NULL) {
826 bzero(sip, sizeof(*sip));
831 * We expect all callers of wait6() to know about WEXITED and
834 error = kern_wait6(td, idtype, id, &status, uap->options, wrup, sip);
836 if (uap->status != NULL && error == 0 && td->td_retval[0] != 0)
837 error = copyout(&status, uap->status, sizeof(status));
838 if (uap->wrusage != NULL && error == 0 && td->td_retval[0] != 0)
839 error = copyout(&wru, uap->wrusage, sizeof(wru));
840 if (uap->info != NULL && error == 0)
841 error = copyout(&si, uap->info, sizeof(si));
846 * Reap the remains of a zombie process and optionally return status and
847 * rusage. Asserts and will release both the proctree_lock and the process
848 * lock as part of its work.
851 proc_reap(struct thread *td, struct proc *p, int *status, int options)
855 sx_assert(&proctree_lock, SA_XLOCKED);
856 PROC_LOCK_ASSERT(p, MA_OWNED);
857 KASSERT(p->p_state == PRS_ZOMBIE, ("proc_reap: !PRS_ZOMBIE"));
859 mtx_spin_wait_unlocked(&p->p_slock);
864 *status = KW_EXITCODE(p->p_xexit, p->p_xsig);
865 if (options & WNOWAIT) {
867 * Only poll, returning the status. Caller does not wish to
868 * release the proc struct just yet.
871 sx_xunlock(&proctree_lock);
876 sigqueue_take(p->p_ksi);
880 * If we got the child via a ptrace 'attach', we need to give it back
883 if (p->p_oppid != p->p_pptr->p_pid) {
885 t = proc_realparent(p);
889 "wait: traced child %d moved back to parent %d", p->p_pid,
891 proc_reparent(p, t, false);
893 pksignal(t, SIGCHLD, p->p_ksi);
895 cv_broadcast(&p->p_pwait);
897 sx_xunlock(&proctree_lock);
903 * Remove other references to this process to ensure we have an
904 * exclusive reference.
906 sx_xlock(&zombproc_lock);
907 LIST_REMOVE(p, p_list); /* off zombproc */
908 sx_xunlock(&zombproc_lock);
909 sx_xlock(PIDHASHLOCK(p->p_pid));
910 LIST_REMOVE(p, p_hash);
911 sx_xunlock(PIDHASHLOCK(p->p_pid));
912 LIST_REMOVE(p, p_sibling);
913 reaper_abandon_children(p, true);
915 proc_id_clear(PROC_ID_PID, p->p_pid);
917 proc_clear_orphan(p);
920 if (p->p_procdesc != NULL)
922 sx_xunlock(&proctree_lock);
925 knlist_detach(p->p_klist);
930 * Removal from allproc list and process group list paired with
931 * PROC_LOCK which was executed during that time should guarantee
932 * nothing can reach this process anymore. As such further locking
935 p->p_xexit = p->p_xsig = 0; /* XXX: why? */
938 ruadd(&q->p_stats->p_cru, &q->p_crux, &p->p_ru, &p->p_rux);
942 * Decrement the count of procs running with this uid.
944 (void)chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0);
947 * Destroy resource accounting information associated with the process.
952 racct_sub(p, RACCT_NPROC, 1);
959 * Free credentials, arguments, and sigacts.
962 proc_set_cred(p, NULL);
963 pargs_drop(p->p_args);
965 sigacts_free(p->p_sigacts);
969 * Do any thread-system specific cleanups.
974 * Give vm and machine-dependent layer a chance to free anything that
975 * cpu_exit couldn't release while still running in process context.
982 KASSERT(FIRST_THREAD_IN_PROC(p),
983 ("proc_reap: no residual thread!"));
984 uma_zfree(proc_zone, p);
985 atomic_add_int(&nprocs, -1);
989 proc_to_reap(struct thread *td, struct proc *p, idtype_t idtype, id_t id,
990 int *status, int options, struct __wrusage *wrusage, siginfo_t *siginfo,
995 sx_assert(&proctree_lock, SA_XLOCKED);
1001 if (p->p_procdesc != NULL) {
1007 if (p->p_pid != (pid_t)id) {
1013 if (p->p_pgid != (pid_t)id) {
1019 if (p->p_session->s_sid != (pid_t)id) {
1025 if (p->p_ucred->cr_uid != (uid_t)id) {
1031 if (p->p_ucred->cr_gid != (gid_t)id) {
1037 if (p->p_ucred->cr_prison->pr_id != (int)id) {
1043 * It seems that the thread structures get zeroed out
1044 * at process exit. This makes it impossible to
1045 * support P_SETID, P_CID or P_CPUID.
1052 if (p_canwait(td, p)) {
1057 if (((options & WEXITED) == 0) && (p->p_state == PRS_ZOMBIE)) {
1063 * This special case handles a kthread spawned by linux_clone
1064 * (see linux_misc.c). The linux_wait4 and linux_waitpid
1065 * functions need to be able to distinguish between waiting
1066 * on a process and waiting on a thread. It is a thread if
1067 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option
1068 * signifies we want to wait for threads and not processes.
1070 if ((p->p_sigparent != SIGCHLD) ^
1071 ((options & WLINUXCLONE) != 0)) {
1076 if (siginfo != NULL) {
1077 bzero(siginfo, sizeof(*siginfo));
1078 siginfo->si_errno = 0;
1081 * SUSv4 requires that the si_signo value is always
1082 * SIGCHLD. Obey it despite the rfork(2) interface
1083 * allows to request other signal for child exit
1086 siginfo->si_signo = SIGCHLD;
1089 * This is still a rough estimate. We will fix the
1090 * cases TRAPPED, STOPPED, and CONTINUED later.
1092 if (WCOREDUMP(p->p_xsig)) {
1093 siginfo->si_code = CLD_DUMPED;
1094 siginfo->si_status = WTERMSIG(p->p_xsig);
1095 } else if (WIFSIGNALED(p->p_xsig)) {
1096 siginfo->si_code = CLD_KILLED;
1097 siginfo->si_status = WTERMSIG(p->p_xsig);
1099 siginfo->si_code = CLD_EXITED;
1100 siginfo->si_status = p->p_xexit;
1103 siginfo->si_pid = p->p_pid;
1104 siginfo->si_uid = p->p_ucred->cr_uid;
1107 * The si_addr field would be useful additional
1108 * detail, but apparently the PC value may be lost
1109 * when we reach this point. bzero() above sets
1110 * siginfo->si_addr to NULL.
1115 * There should be no reason to limit resources usage info to
1116 * exited processes only. A snapshot about any resources used
1117 * by a stopped process may be exactly what is needed.
1119 if (wrusage != NULL) {
1120 rup = &wrusage->wru_self;
1123 calcru(p, &rup->ru_utime, &rup->ru_stime);
1126 rup = &wrusage->wru_children;
1127 *rup = p->p_stats->p_cru;
1128 calccru(p, &rup->ru_utime, &rup->ru_stime);
1131 if (p->p_state == PRS_ZOMBIE && !check_only) {
1132 proc_reap(td, p, status, options);
1139 kern_wait(struct thread *td, pid_t pid, int *status, int options,
1140 struct rusage *rusage)
1142 struct __wrusage wru, *wrup;
1148 * Translate the special pid values into the (idtype, pid)
1149 * pair for kern_wait6. The WAIT_MYPGRP case is handled by
1150 * kern_wait6() on its own.
1152 if (pid == WAIT_ANY) {
1155 } else if (pid < 0) {
1169 * For backward compatibility we implicitly add flags WEXITED
1170 * and WTRAPPED here.
1172 options |= WEXITED | WTRAPPED;
1173 ret = kern_wait6(td, idtype, id, status, options, wrup, NULL);
1175 *rusage = wru.wru_self;
1180 report_alive_proc(struct thread *td, struct proc *p, siginfo_t *siginfo,
1181 int *status, int options, int si_code)
1185 PROC_LOCK_ASSERT(p, MA_OWNED);
1186 sx_assert(&proctree_lock, SA_XLOCKED);
1187 MPASS(si_code == CLD_TRAPPED || si_code == CLD_STOPPED ||
1188 si_code == CLD_CONTINUED);
1190 cont = si_code == CLD_CONTINUED;
1191 if ((options & WNOWAIT) == 0) {
1193 p->p_flag &= ~P_CONTINUED;
1195 p->p_flag |= P_WAITED;
1196 PROC_LOCK(td->td_proc);
1197 sigqueue_take(p->p_ksi);
1198 PROC_UNLOCK(td->td_proc);
1200 sx_xunlock(&proctree_lock);
1201 if (siginfo != NULL) {
1202 siginfo->si_code = si_code;
1203 siginfo->si_status = cont ? SIGCONT : p->p_xsig;
1206 *status = cont ? SIGCONT : W_STOPCODE(p->p_xsig);
1208 td->td_retval[0] = p->p_pid;
1212 kern_wait6(struct thread *td, idtype_t idtype, id_t id, int *status,
1213 int options, struct __wrusage *wrusage, siginfo_t *siginfo)
1217 int error, nfound, ret;
1220 AUDIT_ARG_VALUE((int)idtype); /* XXX - This is likely wrong! */
1221 AUDIT_ARG_PID((pid_t)id); /* XXX - This may be wrong! */
1222 AUDIT_ARG_VALUE(options);
1226 if ((pid_t)id == WAIT_MYPGRP && (idtype == P_PID || idtype == P_PGID)) {
1228 id = (id_t)q->p_pgid;
1233 /* If we don't know the option, just return. */
1234 if ((options & ~(WUNTRACED | WNOHANG | WCONTINUED | WNOWAIT |
1235 WEXITED | WTRAPPED | WLINUXCLONE)) != 0)
1237 if ((options & (WEXITED | WUNTRACED | WCONTINUED | WTRAPPED)) == 0) {
1239 * We will be unable to find any matching processes,
1240 * because there are no known events to look for.
1241 * Prefer to return error instead of blocking
1248 if (q->p_flag & P_STATCHILD) {
1250 q->p_flag &= ~P_STATCHILD;
1253 sx_xlock(&proctree_lock);
1256 LIST_FOREACH(p, &q->p_children, p_sibling) {
1258 ret = proc_to_reap(td, p, idtype, id, status, options,
1259 wrusage, siginfo, 0);
1262 else if (ret != 1) {
1263 td->td_retval[0] = pid;
1268 PROC_LOCK_ASSERT(p, MA_OWNED);
1270 if ((options & WTRAPPED) != 0 &&
1271 (p->p_flag & P_TRACED) != 0) {
1274 ((p->p_flag & (P_STOPPED_TRACE | P_STOPPED_SIG)) &&
1275 p->p_suspcount == p->p_numthreads &&
1276 (p->p_flag & P_WAITED) == 0);
1280 "wait: returning trapped pid %d status %#x "
1281 "(xstat %d) xthread %d",
1282 p->p_pid, W_STOPCODE(p->p_xsig), p->p_xsig,
1283 p->p_xthread != NULL ?
1284 p->p_xthread->td_tid : -1);
1285 report_alive_proc(td, p, siginfo, status,
1286 options, CLD_TRAPPED);
1290 if ((options & WUNTRACED) != 0 &&
1291 (p->p_flag & P_STOPPED_SIG) != 0) {
1293 report = (p->p_suspcount == p->p_numthreads &&
1294 ((p->p_flag & P_WAITED) == 0));
1297 report_alive_proc(td, p, siginfo, status,
1298 options, CLD_STOPPED);
1302 if ((options & WCONTINUED) != 0 &&
1303 (p->p_flag & P_CONTINUED) != 0) {
1304 report_alive_proc(td, p, siginfo, status, options,
1312 * Look in the orphans list too, to allow the parent to
1313 * collect it's child exit status even if child is being
1316 * Debugger detaches from the parent upon successful
1317 * switch-over from parent to child. At this point due to
1318 * re-parenting the parent loses the child to debugger and a
1319 * wait4(2) call would report that it has no children to wait
1320 * for. By maintaining a list of orphans we allow the parent
1321 * to successfully wait until the child becomes a zombie.
1324 LIST_FOREACH(p, &q->p_orphans, p_orphan) {
1325 ret = proc_to_reap(td, p, idtype, id, NULL, options,
1328 KASSERT(ret != -1, ("reaped an orphan (pid %d)",
1329 (int)td->td_retval[0]));
1337 sx_xunlock(&proctree_lock);
1340 if (options & WNOHANG) {
1341 sx_xunlock(&proctree_lock);
1342 td->td_retval[0] = 0;
1346 if (q->p_flag & P_STATCHILD) {
1347 q->p_flag &= ~P_STATCHILD;
1351 sx_xunlock(&proctree_lock);
1352 error = msleep(q, &q->p_mtx, PWAIT | PCATCH | PDROP, "wait", 0);
1359 proc_add_orphan(struct proc *child, struct proc *parent)
1362 sx_assert(&proctree_lock, SX_XLOCKED);
1363 KASSERT((child->p_flag & P_TRACED) != 0,
1364 ("proc_add_orphan: not traced"));
1366 if (LIST_EMPTY(&parent->p_orphans)) {
1367 child->p_treeflag |= P_TREE_FIRST_ORPHAN;
1368 LIST_INSERT_HEAD(&parent->p_orphans, child, p_orphan);
1370 LIST_INSERT_AFTER(LIST_FIRST(&parent->p_orphans),
1373 child->p_treeflag |= P_TREE_ORPHANED;
1377 * Make process 'parent' the new parent of process 'child'.
1378 * Must be called with an exclusive hold of proctree lock.
1381 proc_reparent(struct proc *child, struct proc *parent, bool set_oppid)
1384 sx_assert(&proctree_lock, SX_XLOCKED);
1385 PROC_LOCK_ASSERT(child, MA_OWNED);
1386 if (child->p_pptr == parent)
1389 PROC_LOCK(child->p_pptr);
1390 sigqueue_take(child->p_ksi);
1391 PROC_UNLOCK(child->p_pptr);
1392 LIST_REMOVE(child, p_sibling);
1393 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
1395 proc_clear_orphan(child);
1396 if ((child->p_flag & P_TRACED) != 0) {
1397 proc_add_orphan(child, child->p_pptr);
1400 child->p_pptr = parent;
1402 child->p_oppid = parent->p_pid;