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 reaper_clear(struct proc *p)
156 sx_assert(&proctree_lock, SX_LOCKED);
157 LIST_REMOVE(p, p_reapsibling);
158 if (p->p_reapsubtree == 1)
161 LIST_FOREACH(p1, &p->p_reaper->p_reaplist, p_reapsibling) {
162 if (p1->p_reapsubtree == p->p_reapsubtree) {
168 proc_id_clear(PROC_ID_REAP, p->p_reapsubtree);
172 clear_orphan(struct proc *p)
176 sx_assert(&proctree_lock, SA_XLOCKED);
177 if ((p->p_treeflag & P_TREE_ORPHANED) == 0)
179 if ((p->p_treeflag & P_TREE_FIRST_ORPHAN) != 0) {
180 p1 = LIST_NEXT(p, p_orphan);
182 p1->p_treeflag |= P_TREE_FIRST_ORPHAN;
183 p->p_treeflag &= ~P_TREE_FIRST_ORPHAN;
185 LIST_REMOVE(p, p_orphan);
186 p->p_treeflag &= ~P_TREE_ORPHANED;
190 * exit -- death of process.
193 sys_sys_exit(struct thread *td, struct sys_exit_args *uap)
196 exit1(td, uap->rval, 0);
201 * Exit: deallocate address space and other resources, change proc state to
202 * zombie, and unlink proc from allproc and parent's lists. Save exit status
203 * and rusage for wait(). Check for child processes and orphan them.
206 exit1(struct thread *td, int rval, int signo)
208 struct proc *p, *nq, *q, *t;
210 ksiginfo_t *ksi, *ksi1;
213 mtx_assert(&Giant, MA_NOTOWNED);
214 KASSERT(rval == 0 || signo == 0, ("exit1 rv %d sig %d", rval, signo));
218 * XXX in case we're rebooting we just let init die in order to
219 * work around an unsolved stack overflow seen very late during
220 * shutdown on sparc64 when the gmirror worker process exists.
222 if (p == initproc && rebooting == 0) {
223 printf("init died (signal %d, exit %d)\n", signo, rval);
224 panic("Going nowhere without my init!");
228 * Deref SU mp, since the thread does not return to userspace.
230 td_softdep_cleanup(td);
233 * MUST abort all other threads before proceeding past here.
237 * First check if some other thread or external request got
238 * here before us. If so, act appropriately: exit or suspend.
239 * We must ensure that stop requests are handled before we set
242 thread_suspend_check(0);
243 while (p->p_flag & P_HADTHREADS) {
245 * Kill off the other threads. This requires
246 * some co-operation from other parts of the kernel
247 * so it may not be instantaneous. With this state set
248 * any thread entering the kernel from userspace will
249 * thread_exit() in trap(). Any thread attempting to
250 * sleep will return immediately with EINTR or EWOULDBLOCK
251 * which will hopefully force them to back out to userland
252 * freeing resources as they go. Any thread attempting
253 * to return to userland will thread_exit() from userret().
254 * thread_exit() will unsuspend us when the last of the
255 * other threads exits.
256 * If there is already a thread singler after resumption,
257 * calling thread_single will fail; in that case, we just
258 * re-check all suspension request, the thread should
259 * either be suspended there or exit.
261 if (!thread_single(p, SINGLE_EXIT))
263 * All other activity in this process is now
264 * stopped. Threading support has been turned
269 * Recheck for new stop or suspend requests which
270 * might appear while process lock was dropped in
273 thread_suspend_check(0);
275 KASSERT(p->p_numthreads == 1,
276 ("exit1: proc %p exiting with %d threads", p, p->p_numthreads));
277 racct_sub(p, RACCT_NTHR, 1);
279 /* Let event handler change exit status */
284 * Wakeup anyone in procfs' PIOCWAIT. They should have a hold
285 * on our vmspace, so we should block below until they have
286 * released their reference to us. Note that if they have
287 * requested S_EXIT stops we will block here until they ack
290 _STOPEVENT(p, S_EXIT, 0);
293 * Ignore any pending request to stop due to a stop signal.
294 * Once P_WEXIT is set, future requests will be ignored as
297 p->p_flag &= ~P_STOPPED_SIG;
298 KASSERT(!P_SHOULDSTOP(p), ("exiting process is stopped"));
301 * Note that we are exiting and do another wakeup of anyone in
302 * PIOCWAIT in case they aren't listening for S_EXIT stops or
303 * decided to wait again after we told them we are exiting.
305 p->p_flag |= P_WEXIT;
309 * Wait for any processes that have a hold on our vmspace to
310 * release their reference.
312 while (p->p_lock > 0)
313 msleep(&p->p_lock, &p->p_mtx, PWAIT, "exithold", 0);
316 /* Drain the limit callout while we don't have the proc locked */
317 callout_drain(&p->p_limco);
321 * The Sun BSM exit token contains two components: an exit status as
322 * passed to exit(), and a return value to indicate what sort of exit
323 * it was. The exit status is WEXITSTATUS(rv), but it's not clear
324 * what the return value is.
326 AUDIT_ARG_EXIT(rval, 0);
327 AUDIT_SYSCALL_EXIT(0, td);
330 /* Are we a task leader with peers? */
331 if (p->p_peers != NULL && p == p->p_leader) {
332 mtx_lock(&ppeers_lock);
336 kern_psignal(q, SIGKILL);
340 while (p->p_peers != NULL)
341 msleep(p, &ppeers_lock, PWAIT, "exit1", 0);
342 mtx_unlock(&ppeers_lock);
346 * Check if any loadable modules need anything done at process exit.
347 * E.g. SYSV IPC stuff.
348 * Event handler could change exit status.
349 * XXX what if one of these generates an error?
351 EVENTHANDLER_DIRECT_INVOKE(process_exit, p);
354 * If parent is waiting for us to exit or exec,
355 * P_PPWAIT is set; we will wakeup the parent below.
359 p->p_flag &= ~(P_TRACED | P_PPWAIT | P_PPTRACE);
363 * Stop the real interval timer. If the handler is currently
364 * executing, prevent it from rearming itself and let it finish.
366 if (timevalisset(&p->p_realtimer.it_value) &&
367 _callout_stop_safe(&p->p_itcallout, CS_EXECUTING, NULL) == 0) {
368 timevalclear(&p->p_realtimer.it_interval);
369 msleep(&p->p_itcallout, &p->p_mtx, PWAIT, "ritwait", 0);
370 KASSERT(!timevalisset(&p->p_realtimer.it_value),
371 ("realtime timer is still armed"));
376 umtx_thread_exit(td);
379 * Reset any sigio structures pointing to us as a result of
380 * F_SETOWN with our pid.
382 funsetownlst(&p->p_sigiolst);
385 * If this process has an nlminfo data area (for lockd), release it
387 if (nlminfo_release_p != NULL && p->p_nlminfo != NULL)
388 (*nlminfo_release_p)(p);
391 * Close open files and release open-file table.
397 * If this thread tickled GEOM, we need to wait for the giggling to
398 * stop before we return to userland
400 if (td->td_pflags & TDP_GEOM)
404 * Remove ourself from our leader's peer list and wake our leader.
406 if (p->p_leader->p_peers != NULL) {
407 mtx_lock(&ppeers_lock);
408 if (p->p_leader->p_peers != NULL) {
410 while (q->p_peers != p)
412 q->p_peers = p->p_peers;
415 mtx_unlock(&ppeers_lock);
420 (void)acct_process(td);
426 * Release reference to text vnode
428 if (p->p_textvp != NULL) {
434 * Release our limits structure.
436 lim_free(p->p_limit);
442 * Call machine-dependent code to release any
443 * machine-dependent resources other than the address space.
444 * The address space is released by "vmspace_exitfree(p)" in
449 WITNESS_WARN(WARN_PANIC, NULL, "process (pid %d) exiting", p->p_pid);
452 * Move proc from allproc queue to zombproc.
454 sx_xlock(&allproc_lock);
455 sx_xlock(&zombproc_lock);
456 LIST_REMOVE(p, p_list);
457 LIST_INSERT_HEAD(&zombproc, p, p_list);
458 sx_xunlock(&zombproc_lock);
459 sx_xunlock(&allproc_lock);
461 sx_xlock(&proctree_lock);
464 * Reparent all children processes:
465 * - traced ones to the original parent (or init if we are that parent)
468 q = LIST_FIRST(&p->p_children);
469 if (q != NULL) /* only need this if any child is S_ZOMB */
471 for (; q != NULL; q = nq) {
472 nq = LIST_NEXT(q, p_sibling);
473 ksi = ksiginfo_alloc(TRUE);
475 q->p_sigparent = SIGCHLD;
477 if (!(q->p_flag & P_TRACED)) {
478 proc_reparent(q, q->p_reaper, true);
479 if (q->p_state == PRS_ZOMBIE) {
481 * Inform reaper about the reparented
482 * zombie, since wait(2) has something
483 * new to report. Guarantee queueing
484 * of the SIGCHLD signal, similar to
485 * the _exit() behaviour, by providing
486 * our ksiginfo. Ksi is freed by the
489 if (q->p_ksi == NULL) {
492 ksiginfo_copy(q->p_ksi, ksi);
493 ksi->ksi_flags |= KSI_INS;
497 PROC_LOCK(q->p_reaper);
498 pksignal(q->p_reaper, SIGCHLD, ksi1);
499 PROC_UNLOCK(q->p_reaper);
500 } else if (q->p_pdeathsig > 0) {
502 * The child asked to received a signal
505 kern_psignal(q, q->p_pdeathsig);
509 * Traced processes are killed since their existence
510 * means someone is screwing up.
512 t = proc_realparent(q);
514 proc_reparent(q, q->p_reaper, true);
517 proc_reparent(q, t, true);
521 * Since q was found on our children list, the
522 * proc_reparent() call moved q to the orphan
523 * list due to present P_TRACED flag. Clear
524 * orphan link for q now while q is locked.
527 q->p_flag &= ~(P_TRACED | P_STOPPED_TRACE);
528 q->p_flag2 &= ~P2_PTRACE_FSTP;
530 FOREACH_THREAD_IN_PROC(q, tdt) {
531 tdt->td_dbgflags &= ~(TDB_SUSPEND | TDB_XSIG |
534 kern_psignal(q, SIGKILL);
542 * Also get rid of our orphans.
544 while ((q = LIST_FIRST(&p->p_orphans)) != NULL) {
546 KASSERT(q->p_oppid == p->p_pid,
547 ("orphan %p of %p has unexpected oppid %d", q, p,
549 q->p_oppid = q->p_reaper->p_pid;
552 * If we are the real parent of this process
553 * but it has been reparented to a debugger, then
554 * check if it asked for a signal when we exit.
556 if (q->p_pdeathsig > 0)
557 kern_psignal(q, q->p_pdeathsig);
558 CTR2(KTR_PTRACE, "exit: pid %d, clearing orphan %d", p->p_pid,
565 if (SDT_PROBES_ENABLED()) {
566 int reason = CLD_EXITED;
567 if (WCOREDUMP(signo))
569 else if (WIFSIGNALED(signo))
571 SDT_PROBE1(proc, , , exit, reason);
575 /* Save exit status. */
581 * Tell the DTrace fasttrap provider about the exit if it
582 * has declared an interest.
584 if (dtrace_fasttrap_exit)
585 dtrace_fasttrap_exit(p);
589 * Notify interested parties of our demise.
591 KNOTE_LOCKED(p->p_klist, NOTE_EXIT);
594 * If this is a process with a descriptor, we may not need to deliver
595 * a signal to the parent. proctree_lock is held over
596 * procdesc_exit() to serialize concurrent calls to close() and
600 if (p->p_procdesc == NULL || procdesc_exit(p)) {
602 * Notify parent that we're gone. If parent has the
603 * PS_NOCLDWAIT flag set, or if the handler is set to SIG_IGN,
604 * notify process 1 instead (and hope it will handle this
607 PROC_LOCK(p->p_pptr);
608 mtx_lock(&p->p_pptr->p_sigacts->ps_mtx);
609 if (p->p_pptr->p_sigacts->ps_flag &
610 (PS_NOCLDWAIT | PS_CLDSIGIGN)) {
613 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
616 proc_reparent(p, p->p_reaper, true);
617 p->p_sigparent = SIGCHLD;
618 PROC_LOCK(p->p_pptr);
621 * Notify parent, so in case he was wait(2)ing or
622 * executing waitpid(2) with our pid, he will
627 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
629 if (p->p_pptr == p->p_reaper || p->p_pptr == initproc) {
631 } else if (p->p_sigparent != 0) {
632 if (p->p_sigparent == SIGCHLD) {
634 } else { /* LINUX thread */
639 PROC_LOCK(p->p_pptr);
640 sx_xunlock(&proctree_lock);
642 if (signal_parent == 1) {
644 } else if (signal_parent == 2) {
645 kern_psignal(p->p_pptr, p->p_sigparent);
648 /* Tell the prison that we are gone. */
649 prison_proc_free(p->p_ucred->cr_prison);
652 * The state PRS_ZOMBIE prevents other proesses from sending
653 * signal to the process, to avoid memory leak, we free memory
654 * for signal queue at the time when the state is set.
656 sigqueue_flush(&p->p_sigqueue);
657 sigqueue_flush(&td->td_sigqueue);
660 * We have to wait until after acquiring all locks before
661 * changing p_state. We need to avoid all possible context
662 * switches (including ones from blocking on a mutex) while
663 * marked as a zombie. We also have to set the zombie state
664 * before we release the parent process' proc lock to avoid
665 * a lost wakeup. So, we first call wakeup, then we grab the
666 * sched lock, update the state, and release the parent process'
670 cv_broadcast(&p->p_pwait);
671 sched_exit(p->p_pptr, td);
673 p->p_state = PRS_ZOMBIE;
674 PROC_UNLOCK(p->p_pptr);
677 * Save our children's rusage information in our exit rusage.
680 ruadd(&p->p_ru, &p->p_rux, &p->p_stats->p_cru, &p->p_crux);
684 * Make sure the scheduler takes this thread out of its tables etc.
685 * This will also release this thread's reference to the ucred.
686 * Other thread parts to release include pcb bits and such.
692 #ifndef _SYS_SYSPROTO_H_
701 sys_abort2(struct thread *td, struct abort2_args *uap)
703 struct proc *p = td->td_proc;
709 * Do it right now so we can log either proper call of abort2(), or
710 * note, that invalid argument was passed. 512 is big enough to
711 * handle 16 arguments' descriptions with additional comments.
713 sb = sbuf_new(NULL, NULL, 512, SBUF_FIXEDLEN);
715 sbuf_printf(sb, "%s(pid %d uid %d) aborted: ",
716 p->p_comm, p->p_pid, td->td_ucred->cr_uid);
718 * Since we can't return from abort2(), send SIGKILL in cases, where
719 * abort2() was called improperly
722 /* Prevent from DoSes from user-space. */
723 if (uap->nargs < 0 || uap->nargs > 16)
725 if (uap->nargs > 0) {
726 if (uap->args == NULL)
728 error = copyin(uap->args, uargs, uap->nargs * sizeof(void *));
733 * Limit size of 'reason' string to 128. Will fit even when
734 * maximal number of arguments was chosen to be logged.
736 if (uap->why != NULL) {
737 error = sbuf_copyin(sb, uap->why, 128);
741 sbuf_printf(sb, "(null)");
743 if (uap->nargs > 0) {
744 sbuf_printf(sb, "(");
745 for (i = 0;i < uap->nargs; i++)
746 sbuf_printf(sb, "%s%p", i == 0 ? "" : ", ", uargs[i]);
747 sbuf_printf(sb, ")");
750 * Final stage: arguments were proper, string has been
751 * successfully copied from userspace, and copying pointers
752 * from user-space succeed.
756 if (sig == SIGKILL) {
758 sbuf_printf(sb, " (Reason text inaccessible)");
762 log(LOG_INFO, "%s", sbuf_data(sb));
771 * The dirty work is handled by kern_wait().
774 owait(struct thread *td, struct owait_args *uap __unused)
778 error = kern_wait(td, WAIT_ANY, &status, 0, NULL);
780 td->td_retval[1] = status;
783 #endif /* COMPAT_43 */
786 * The dirty work is handled by kern_wait().
789 sys_wait4(struct thread *td, struct wait4_args *uap)
791 struct rusage ru, *rup;
794 if (uap->rusage != NULL)
798 error = kern_wait(td, uap->pid, &status, uap->options, rup);
799 if (uap->status != NULL && error == 0 && td->td_retval[0] != 0)
800 error = copyout(&status, uap->status, sizeof(status));
801 if (uap->rusage != NULL && error == 0 && td->td_retval[0] != 0)
802 error = copyout(&ru, uap->rusage, sizeof(struct rusage));
807 sys_wait6(struct thread *td, struct wait6_args *uap)
809 struct __wrusage wru, *wrup;
815 idtype = uap->idtype;
818 if (uap->wrusage != NULL)
823 if (uap->info != NULL) {
825 bzero(sip, sizeof(*sip));
830 * We expect all callers of wait6() to know about WEXITED and
833 error = kern_wait6(td, idtype, id, &status, uap->options, wrup, sip);
835 if (uap->status != NULL && error == 0 && td->td_retval[0] != 0)
836 error = copyout(&status, uap->status, sizeof(status));
837 if (uap->wrusage != NULL && error == 0 && td->td_retval[0] != 0)
838 error = copyout(&wru, uap->wrusage, sizeof(wru));
839 if (uap->info != NULL && error == 0)
840 error = copyout(&si, uap->info, sizeof(si));
845 * Reap the remains of a zombie process and optionally return status and
846 * rusage. Asserts and will release both the proctree_lock and the process
847 * lock as part of its work.
850 proc_reap(struct thread *td, struct proc *p, int *status, int options)
854 sx_assert(&proctree_lock, SA_XLOCKED);
855 PROC_LOCK_ASSERT(p, MA_OWNED);
856 KASSERT(p->p_state == PRS_ZOMBIE, ("proc_reap: !PRS_ZOMBIE"));
858 mtx_spin_wait_unlocked(&p->p_slock);
863 *status = KW_EXITCODE(p->p_xexit, p->p_xsig);
864 if (options & WNOWAIT) {
866 * Only poll, returning the status. Caller does not wish to
867 * release the proc struct just yet.
870 sx_xunlock(&proctree_lock);
875 sigqueue_take(p->p_ksi);
879 * If we got the child via a ptrace 'attach', we need to give it back
882 if (p->p_oppid != p->p_pptr->p_pid) {
884 t = proc_realparent(p);
888 "wait: traced child %d moved back to parent %d", p->p_pid,
890 proc_reparent(p, t, false);
892 pksignal(t, SIGCHLD, p->p_ksi);
894 cv_broadcast(&p->p_pwait);
896 sx_xunlock(&proctree_lock);
902 * Remove other references to this process to ensure we have an
903 * exclusive reference.
905 sx_xlock(&zombproc_lock);
906 LIST_REMOVE(p, p_list); /* off zombproc */
907 sx_xunlock(&zombproc_lock);
908 sx_xlock(PIDHASHLOCK(p->p_pid));
909 LIST_REMOVE(p, p_hash);
910 sx_xunlock(PIDHASHLOCK(p->p_pid));
911 LIST_REMOVE(p, p_sibling);
912 reaper_abandon_children(p, true);
914 proc_id_clear(PROC_ID_PID, p->p_pid);
919 if (p->p_procdesc != NULL)
921 sx_xunlock(&proctree_lock);
924 knlist_detach(p->p_klist);
929 * Removal from allproc list and process group list paired with
930 * PROC_LOCK which was executed during that time should guarantee
931 * nothing can reach this process anymore. As such further locking
934 p->p_xexit = p->p_xsig = 0; /* XXX: why? */
937 ruadd(&q->p_stats->p_cru, &q->p_crux, &p->p_ru, &p->p_rux);
941 * Decrement the count of procs running with this uid.
943 (void)chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0);
946 * Destroy resource accounting information associated with the process.
951 racct_sub(p, RACCT_NPROC, 1);
958 * Free credentials, arguments, and sigacts.
961 proc_set_cred(p, NULL);
962 pargs_drop(p->p_args);
964 sigacts_free(p->p_sigacts);
968 * Do any thread-system specific cleanups.
973 * Give vm and machine-dependent layer a chance to free anything that
974 * cpu_exit couldn't release while still running in process context.
981 KASSERT(FIRST_THREAD_IN_PROC(p),
982 ("proc_reap: no residual thread!"));
983 uma_zfree(proc_zone, p);
984 atomic_add_int(&nprocs, -1);
988 proc_to_reap(struct thread *td, struct proc *p, idtype_t idtype, id_t id,
989 int *status, int options, struct __wrusage *wrusage, siginfo_t *siginfo,
994 sx_assert(&proctree_lock, SA_XLOCKED);
1000 if (p->p_procdesc != NULL) {
1006 if (p->p_pid != (pid_t)id) {
1012 if (p->p_pgid != (pid_t)id) {
1018 if (p->p_session->s_sid != (pid_t)id) {
1024 if (p->p_ucred->cr_uid != (uid_t)id) {
1030 if (p->p_ucred->cr_gid != (gid_t)id) {
1036 if (p->p_ucred->cr_prison->pr_id != (int)id) {
1042 * It seems that the thread structures get zeroed out
1043 * at process exit. This makes it impossible to
1044 * support P_SETID, P_CID or P_CPUID.
1051 if (p_canwait(td, p)) {
1056 if (((options & WEXITED) == 0) && (p->p_state == PRS_ZOMBIE)) {
1062 * This special case handles a kthread spawned by linux_clone
1063 * (see linux_misc.c). The linux_wait4 and linux_waitpid
1064 * functions need to be able to distinguish between waiting
1065 * on a process and waiting on a thread. It is a thread if
1066 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option
1067 * signifies we want to wait for threads and not processes.
1069 if ((p->p_sigparent != SIGCHLD) ^
1070 ((options & WLINUXCLONE) != 0)) {
1075 if (siginfo != NULL) {
1076 bzero(siginfo, sizeof(*siginfo));
1077 siginfo->si_errno = 0;
1080 * SUSv4 requires that the si_signo value is always
1081 * SIGCHLD. Obey it despite the rfork(2) interface
1082 * allows to request other signal for child exit
1085 siginfo->si_signo = SIGCHLD;
1088 * This is still a rough estimate. We will fix the
1089 * cases TRAPPED, STOPPED, and CONTINUED later.
1091 if (WCOREDUMP(p->p_xsig)) {
1092 siginfo->si_code = CLD_DUMPED;
1093 siginfo->si_status = WTERMSIG(p->p_xsig);
1094 } else if (WIFSIGNALED(p->p_xsig)) {
1095 siginfo->si_code = CLD_KILLED;
1096 siginfo->si_status = WTERMSIG(p->p_xsig);
1098 siginfo->si_code = CLD_EXITED;
1099 siginfo->si_status = p->p_xexit;
1102 siginfo->si_pid = p->p_pid;
1103 siginfo->si_uid = p->p_ucred->cr_uid;
1106 * The si_addr field would be useful additional
1107 * detail, but apparently the PC value may be lost
1108 * when we reach this point. bzero() above sets
1109 * siginfo->si_addr to NULL.
1114 * There should be no reason to limit resources usage info to
1115 * exited processes only. A snapshot about any resources used
1116 * by a stopped process may be exactly what is needed.
1118 if (wrusage != NULL) {
1119 rup = &wrusage->wru_self;
1122 calcru(p, &rup->ru_utime, &rup->ru_stime);
1125 rup = &wrusage->wru_children;
1126 *rup = p->p_stats->p_cru;
1127 calccru(p, &rup->ru_utime, &rup->ru_stime);
1130 if (p->p_state == PRS_ZOMBIE && !check_only) {
1131 proc_reap(td, p, status, options);
1138 kern_wait(struct thread *td, pid_t pid, int *status, int options,
1139 struct rusage *rusage)
1141 struct __wrusage wru, *wrup;
1147 * Translate the special pid values into the (idtype, pid)
1148 * pair for kern_wait6. The WAIT_MYPGRP case is handled by
1149 * kern_wait6() on its own.
1151 if (pid == WAIT_ANY) {
1154 } else if (pid < 0) {
1168 * For backward compatibility we implicitly add flags WEXITED
1169 * and WTRAPPED here.
1171 options |= WEXITED | WTRAPPED;
1172 ret = kern_wait6(td, idtype, id, status, options, wrup, NULL);
1174 *rusage = wru.wru_self;
1179 report_alive_proc(struct thread *td, struct proc *p, siginfo_t *siginfo,
1180 int *status, int options, int si_code)
1184 PROC_LOCK_ASSERT(p, MA_OWNED);
1185 sx_assert(&proctree_lock, SA_XLOCKED);
1186 MPASS(si_code == CLD_TRAPPED || si_code == CLD_STOPPED ||
1187 si_code == CLD_CONTINUED);
1189 cont = si_code == CLD_CONTINUED;
1190 if ((options & WNOWAIT) == 0) {
1192 p->p_flag &= ~P_CONTINUED;
1194 p->p_flag |= P_WAITED;
1195 PROC_LOCK(td->td_proc);
1196 sigqueue_take(p->p_ksi);
1197 PROC_UNLOCK(td->td_proc);
1199 sx_xunlock(&proctree_lock);
1200 if (siginfo != NULL) {
1201 siginfo->si_code = si_code;
1202 siginfo->si_status = cont ? SIGCONT : p->p_xsig;
1205 *status = cont ? SIGCONT : W_STOPCODE(p->p_xsig);
1207 td->td_retval[0] = p->p_pid;
1211 kern_wait6(struct thread *td, idtype_t idtype, id_t id, int *status,
1212 int options, struct __wrusage *wrusage, siginfo_t *siginfo)
1216 int error, nfound, ret;
1219 AUDIT_ARG_VALUE((int)idtype); /* XXX - This is likely wrong! */
1220 AUDIT_ARG_PID((pid_t)id); /* XXX - This may be wrong! */
1221 AUDIT_ARG_VALUE(options);
1225 if ((pid_t)id == WAIT_MYPGRP && (idtype == P_PID || idtype == P_PGID)) {
1227 id = (id_t)q->p_pgid;
1232 /* If we don't know the option, just return. */
1233 if ((options & ~(WUNTRACED | WNOHANG | WCONTINUED | WNOWAIT |
1234 WEXITED | WTRAPPED | WLINUXCLONE)) != 0)
1236 if ((options & (WEXITED | WUNTRACED | WCONTINUED | WTRAPPED)) == 0) {
1238 * We will be unable to find any matching processes,
1239 * because there are no known events to look for.
1240 * Prefer to return error instead of blocking
1247 if (q->p_flag & P_STATCHILD) {
1249 q->p_flag &= ~P_STATCHILD;
1252 sx_xlock(&proctree_lock);
1255 LIST_FOREACH(p, &q->p_children, p_sibling) {
1257 ret = proc_to_reap(td, p, idtype, id, status, options,
1258 wrusage, siginfo, 0);
1261 else if (ret != 1) {
1262 td->td_retval[0] = pid;
1267 PROC_LOCK_ASSERT(p, MA_OWNED);
1269 if ((options & WTRAPPED) != 0 &&
1270 (p->p_flag & P_TRACED) != 0) {
1273 ((p->p_flag & (P_STOPPED_TRACE | P_STOPPED_SIG)) &&
1274 p->p_suspcount == p->p_numthreads &&
1275 (p->p_flag & P_WAITED) == 0);
1279 "wait: returning trapped pid %d status %#x "
1280 "(xstat %d) xthread %d",
1281 p->p_pid, W_STOPCODE(p->p_xsig), p->p_xsig,
1282 p->p_xthread != NULL ?
1283 p->p_xthread->td_tid : -1);
1284 report_alive_proc(td, p, siginfo, status,
1285 options, CLD_TRAPPED);
1289 if ((options & WUNTRACED) != 0 &&
1290 (p->p_flag & P_STOPPED_SIG) != 0) {
1292 report = (p->p_suspcount == p->p_numthreads &&
1293 ((p->p_flag & P_WAITED) == 0));
1296 report_alive_proc(td, p, siginfo, status,
1297 options, CLD_STOPPED);
1301 if ((options & WCONTINUED) != 0 &&
1302 (p->p_flag & P_CONTINUED) != 0) {
1303 report_alive_proc(td, p, siginfo, status, options,
1311 * Look in the orphans list too, to allow the parent to
1312 * collect it's child exit status even if child is being
1315 * Debugger detaches from the parent upon successful
1316 * switch-over from parent to child. At this point due to
1317 * re-parenting the parent loses the child to debugger and a
1318 * wait4(2) call would report that it has no children to wait
1319 * for. By maintaining a list of orphans we allow the parent
1320 * to successfully wait until the child becomes a zombie.
1323 LIST_FOREACH(p, &q->p_orphans, p_orphan) {
1324 ret = proc_to_reap(td, p, idtype, id, NULL, options,
1327 KASSERT(ret != -1, ("reaped an orphan (pid %d)",
1328 (int)td->td_retval[0]));
1336 sx_xunlock(&proctree_lock);
1339 if (options & WNOHANG) {
1340 sx_xunlock(&proctree_lock);
1341 td->td_retval[0] = 0;
1345 if (q->p_flag & P_STATCHILD) {
1346 q->p_flag &= ~P_STATCHILD;
1350 sx_xunlock(&proctree_lock);
1351 error = msleep(q, &q->p_mtx, PWAIT | PCATCH | PDROP, "wait", 0);
1358 * Make process 'parent' the new parent of process 'child'.
1359 * Must be called with an exclusive hold of proctree lock.
1362 proc_reparent(struct proc *child, struct proc *parent, bool set_oppid)
1365 sx_assert(&proctree_lock, SX_XLOCKED);
1366 PROC_LOCK_ASSERT(child, MA_OWNED);
1367 if (child->p_pptr == parent)
1370 PROC_LOCK(child->p_pptr);
1371 sigqueue_take(child->p_ksi);
1372 PROC_UNLOCK(child->p_pptr);
1373 LIST_REMOVE(child, p_sibling);
1374 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
1376 clear_orphan(child);
1377 if (child->p_flag & P_TRACED) {
1378 if (LIST_EMPTY(&child->p_pptr->p_orphans)) {
1379 child->p_treeflag |= P_TREE_FIRST_ORPHAN;
1380 LIST_INSERT_HEAD(&child->p_pptr->p_orphans, child,
1383 LIST_INSERT_AFTER(LIST_FIRST(&child->p_pptr->p_orphans),
1386 child->p_treeflag |= P_TREE_ORPHANED;
1389 child->p_pptr = parent;
1391 child->p_oppid = parent->p_pid;