2 * Copyright (c) 1982, 1986, 1989, 1991, 1993
3 * The Regents of the University of California. All rights reserved.
4 * (c) UNIX System Laboratories, Inc.
5 * All or some portions of this file are derived from material licensed
6 * to the University of California by American Telephone and Telegraph
7 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8 * the permission of UNIX System Laboratories, Inc.
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 4. Neither the name of the University nor the names of its contributors
19 * may be used to endorse or promote products derived from this software
20 * without specific prior written permission.
22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * @(#)kern_exit.c 8.7 (Berkeley) 2/12/94
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD$");
40 #include "opt_compat.h"
41 #include "opt_ktrace.h"
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/sysproto.h>
46 #include <sys/capsicum.h>
47 #include <sys/eventhandler.h>
48 #include <sys/kernel.h>
49 #include <sys/malloc.h>
51 #include <sys/mutex.h>
53 #include <sys/procdesc.h>
54 #include <sys/pioctl.h>
58 #include <sys/vmmeter.h>
59 #include <sys/vnode.h>
60 #include <sys/racct.h>
61 #include <sys/resourcevar.h>
63 #include <sys/signalvar.h>
64 #include <sys/sched.h>
66 #include <sys/syscallsubr.h>
67 #include <sys/syslog.h>
68 #include <sys/ptrace.h>
69 #include <sys/acct.h> /* for acct_process() function prototype */
70 #include <sys/filedesc.h>
76 #include <sys/ktrace.h>
79 #include <security/audit/audit.h>
80 #include <security/mac/mac_framework.h>
83 #include <vm/vm_extern.h>
84 #include <vm/vm_param.h>
86 #include <vm/vm_map.h>
87 #include <vm/vm_page.h>
89 #include <vm/vm_domain.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, kernel, , exit, "int");
99 /* Hook for NFS teardown procedure. */
100 void (*nlminfo_release_p)(struct proc *p);
103 proc_realparent(struct proc *child)
105 struct proc *p, *parent;
107 sx_assert(&proctree_lock, SX_LOCKED);
108 if ((child->p_treeflag & P_TREE_ORPHANED) == 0) {
109 if (child->p_oppid == 0 ||
110 child->p_pptr->p_pid == child->p_oppid)
111 parent = child->p_pptr;
116 for (p = child; (p->p_treeflag & P_TREE_FIRST_ORPHAN) == 0;) {
117 /* Cannot use LIST_PREV(), since the list head is not known. */
118 p = __containerof(p->p_orphan.le_prev, struct proc,
120 KASSERT((p->p_treeflag & P_TREE_ORPHANED) != 0,
121 ("missing P_ORPHAN %p", p));
123 parent = __containerof(p->p_orphan.le_prev, struct proc,
129 reaper_abandon_children(struct proc *p, bool exiting)
131 struct proc *p1, *p2, *ptmp;
133 sx_assert(&proctree_lock, SX_LOCKED);
134 KASSERT(p != initproc, ("reaper_abandon_children for initproc"));
135 if ((p->p_treeflag & P_TREE_REAPER) == 0)
138 LIST_FOREACH_SAFE(p2, &p->p_reaplist, p_reapsibling, ptmp) {
139 LIST_REMOVE(p2, p_reapsibling);
141 p2->p_reapsubtree = p->p_reapsubtree;
142 LIST_INSERT_HEAD(&p1->p_reaplist, p2, p_reapsibling);
143 if (exiting && p2->p_pptr == p) {
145 proc_reparent(p2, p1);
149 KASSERT(LIST_EMPTY(&p->p_reaplist), ("p_reaplist not empty"));
150 p->p_treeflag &= ~P_TREE_REAPER;
154 clear_orphan(struct proc *p)
158 sx_assert(&proctree_lock, SA_XLOCKED);
159 if ((p->p_treeflag & P_TREE_ORPHANED) == 0)
161 if ((p->p_treeflag & P_TREE_FIRST_ORPHAN) != 0) {
162 p1 = LIST_NEXT(p, p_orphan);
164 p1->p_treeflag |= P_TREE_FIRST_ORPHAN;
165 p->p_treeflag &= ~P_TREE_FIRST_ORPHAN;
167 LIST_REMOVE(p, p_orphan);
168 p->p_treeflag &= ~P_TREE_ORPHANED;
172 * exit -- death of process.
175 sys_sys_exit(struct thread *td, struct sys_exit_args *uap)
178 exit1(td, uap->rval, 0);
183 * Exit: deallocate address space and other resources, change proc state to
184 * zombie, and unlink proc from allproc and parent's lists. Save exit status
185 * and rusage for wait(). Check for child processes and orphan them.
188 exit1(struct thread *td, int rval, int signo)
190 struct proc *p, *nq, *q, *t;
192 struct vnode *ttyvp = NULL;
194 mtx_assert(&Giant, MA_NOTOWNED);
195 KASSERT(rval == 0 || signo == 0, ("exit1 rv %d sig %d", rval, signo));
199 * XXX in case we're rebooting we just let init die in order to
200 * work around an unsolved stack overflow seen very late during
201 * shutdown on sparc64 when the gmirror worker process exists.
203 if (p == initproc && rebooting == 0) {
204 printf("init died (signal %d, exit %d)\n", signo, rval);
205 panic("Going nowhere without my init!");
209 * Deref SU mp, since the thread does not return to userspace.
211 if (softdep_ast_cleanup != NULL)
212 softdep_ast_cleanup();
215 * MUST abort all other threads before proceeding past here.
219 * First check if some other thread or external request got
220 * here before us. If so, act appropriately: exit or suspend.
221 * We must ensure that stop requests are handled before we set
224 thread_suspend_check(0);
225 while (p->p_flag & P_HADTHREADS) {
227 * Kill off the other threads. This requires
228 * some co-operation from other parts of the kernel
229 * so it may not be instantaneous. With this state set
230 * any thread entering the kernel from userspace will
231 * thread_exit() in trap(). Any thread attempting to
232 * sleep will return immediately with EINTR or EWOULDBLOCK
233 * which will hopefully force them to back out to userland
234 * freeing resources as they go. Any thread attempting
235 * to return to userland will thread_exit() from userret().
236 * thread_exit() will unsuspend us when the last of the
237 * other threads exits.
238 * If there is already a thread singler after resumption,
239 * calling thread_single will fail; in that case, we just
240 * re-check all suspension request, the thread should
241 * either be suspended there or exit.
243 if (!thread_single(p, SINGLE_EXIT))
245 * All other activity in this process is now
246 * stopped. Threading support has been turned
251 * Recheck for new stop or suspend requests which
252 * might appear while process lock was dropped in
255 thread_suspend_check(0);
257 KASSERT(p->p_numthreads == 1,
258 ("exit1: proc %p exiting with %d threads", p, p->p_numthreads));
259 racct_sub(p, RACCT_NTHR, 1);
261 /* Let event handler change exit status */
266 * Wakeup anyone in procfs' PIOCWAIT. They should have a hold
267 * on our vmspace, so we should block below until they have
268 * released their reference to us. Note that if they have
269 * requested S_EXIT stops we will block here until they ack
272 _STOPEVENT(p, S_EXIT, 0);
275 * Ignore any pending request to stop due to a stop signal.
276 * Once P_WEXIT is set, future requests will be ignored as
279 p->p_flag &= ~P_STOPPED_SIG;
280 KASSERT(!P_SHOULDSTOP(p), ("exiting process is stopped"));
283 * Note that we are exiting and do another wakeup of anyone in
284 * PIOCWAIT in case they aren't listening for S_EXIT stops or
285 * decided to wait again after we told them we are exiting.
287 p->p_flag |= P_WEXIT;
291 * Wait for any processes that have a hold on our vmspace to
292 * release their reference.
294 while (p->p_lock > 0)
295 msleep(&p->p_lock, &p->p_mtx, PWAIT, "exithold", 0);
298 /* Drain the limit callout while we don't have the proc locked */
299 callout_drain(&p->p_limco);
303 * The Sun BSM exit token contains two components: an exit status as
304 * passed to exit(), and a return value to indicate what sort of exit
305 * it was. The exit status is WEXITSTATUS(rv), but it's not clear
306 * what the return value is.
308 AUDIT_ARG_EXIT(rval, 0);
309 AUDIT_SYSCALL_EXIT(0, td);
312 /* Are we a task leader with peers? */
313 if (p->p_peers != NULL && p == p->p_leader) {
314 mtx_lock(&ppeers_lock);
318 kern_psignal(q, SIGKILL);
322 while (p->p_peers != NULL)
323 msleep(p, &ppeers_lock, PWAIT, "exit1", 0);
324 mtx_unlock(&ppeers_lock);
328 * Check if any loadable modules need anything done at process exit.
329 * E.g. SYSV IPC stuff.
330 * Event handler could change exit status.
331 * XXX what if one of these generates an error?
333 EVENTHANDLER_INVOKE(process_exit, p);
336 * If parent is waiting for us to exit or exec,
337 * P_PPWAIT is set; we will wakeup the parent below.
341 p->p_flag &= ~(P_TRACED | P_PPWAIT | P_PPTRACE);
344 * Stop the real interval timer. If the handler is currently
345 * executing, prevent it from rearming itself and let it finish.
347 if (timevalisset(&p->p_realtimer.it_value) &&
348 callout_stop(&p->p_itcallout) == 0) {
349 timevalclear(&p->p_realtimer.it_interval);
350 msleep(&p->p_itcallout, &p->p_mtx, PWAIT, "ritwait", 0);
351 KASSERT(!timevalisset(&p->p_realtimer.it_value),
352 ("realtime timer is still armed"));
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 sx_xlock(&proctree_lock);
399 if (SESS_LEADER(p)) {
400 struct session *sp = p->p_session;
404 * s_ttyp is not zero'd; we use this to indicate that
405 * the session once had a controlling terminal. (for
406 * logging and informational purposes)
417 * Signal foreground pgrp and revoke access to
418 * controlling terminal if it has not been revoked
421 * Because the TTY may have been revoked in the mean
422 * time and could already have a new session associated
423 * with it, make sure we don't send a SIGHUP to a
424 * foreground process group that does not belong to this
430 if (tp->t_session == sp)
431 tty_signal_pgrp(tp, SIGHUP);
436 sx_xunlock(&proctree_lock);
437 if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
438 VOP_REVOKE(ttyvp, REVOKEALL);
439 VOP_UNLOCK(ttyvp, 0);
441 sx_xlock(&proctree_lock);
444 fixjobc(p, p->p_pgrp, 0);
445 sx_xunlock(&proctree_lock);
446 (void)acct_process(td);
448 /* Release the TTY now we've unlocked everything. */
455 * Release reference to text vnode
457 if (p->p_textvp != NULL) {
463 * Release our limits structure.
465 lim_free(p->p_limit);
471 * Remove proc from allproc queue and pidhash chain.
472 * Place onto zombproc. Unlink from parent's child list.
474 sx_xlock(&allproc_lock);
475 LIST_REMOVE(p, p_list);
476 LIST_INSERT_HEAD(&zombproc, p, p_list);
477 LIST_REMOVE(p, p_hash);
478 sx_xunlock(&allproc_lock);
481 * Call machine-dependent code to release any
482 * machine-dependent resources other than the address space.
483 * The address space is released by "vmspace_exitfree(p)" in
488 WITNESS_WARN(WARN_PANIC, NULL, "process (pid %d) exiting", p->p_pid);
491 * Reparent all children processes:
492 * - traced ones to the original parent (or init if we are that parent)
495 sx_xlock(&proctree_lock);
496 q = LIST_FIRST(&p->p_children);
497 if (q != NULL) /* only need this if any child is S_ZOMB */
499 for (; q != NULL; q = nq) {
500 nq = LIST_NEXT(q, p_sibling);
502 q->p_sigparent = SIGCHLD;
504 if (!(q->p_flag & P_TRACED)) {
505 proc_reparent(q, q->p_reaper);
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);
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 FOREACH_THREAD_IN_PROC(q, tdt)
528 tdt->td_dbgflags &= ~TDB_SUSPEND;
529 kern_psignal(q, SIGKILL);
535 * Also get rid of our orphans.
537 while ((q = LIST_FIRST(&p->p_orphans)) != NULL) {
539 CTR2(KTR_PTRACE, "exit: pid %d, clearing orphan %d", p->p_pid,
545 /* Save exit status. */
549 /* Tell the prison that we are gone. */
550 prison_proc_free(p->p_ucred->cr_prison);
554 * Tell the DTrace fasttrap provider about the exit if it
555 * has declared an interest.
557 if (dtrace_fasttrap_exit)
558 dtrace_fasttrap_exit(p);
562 * Notify interested parties of our demise.
564 KNOTE_LOCKED(&p->p_klist, NOTE_EXIT);
567 int reason = CLD_EXITED;
568 if (WCOREDUMP(signo))
570 else if (WIFSIGNALED(signo))
572 SDT_PROBE1(proc, kernel, , exit, reason);
576 * Just delete all entries in the p_klist. At this point we won't
577 * report any more events, and there are nasty race conditions that
578 * can beat us if we don't.
580 knlist_clear(&p->p_klist, 1);
583 * If this is a process with a descriptor, we may not need to deliver
584 * a signal to the parent. proctree_lock is held over
585 * procdesc_exit() to serialize concurrent calls to close() and
588 if (p->p_procdesc == NULL || procdesc_exit(p)) {
590 * Notify parent that we're gone. If parent has the
591 * PS_NOCLDWAIT flag set, or if the handler is set to SIG_IGN,
592 * notify process 1 instead (and hope it will handle this
595 PROC_LOCK(p->p_pptr);
596 mtx_lock(&p->p_pptr->p_sigacts->ps_mtx);
597 if (p->p_pptr->p_sigacts->ps_flag &
598 (PS_NOCLDWAIT | PS_CLDSIGIGN)) {
601 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
604 proc_reparent(p, p->p_reaper);
605 p->p_sigparent = SIGCHLD;
606 PROC_LOCK(p->p_pptr);
609 * Notify parent, so in case he was wait(2)ing or
610 * executing waitpid(2) with our pid, he will
615 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
617 if (p->p_pptr == p->p_reaper || p->p_pptr == initproc)
619 else if (p->p_sigparent != 0) {
620 if (p->p_sigparent == SIGCHLD)
622 else /* LINUX thread */
623 kern_psignal(p->p_pptr, p->p_sigparent);
626 PROC_LOCK(p->p_pptr);
627 sx_xunlock(&proctree_lock);
630 * The state PRS_ZOMBIE prevents other proesses from sending
631 * signal to the process, to avoid memory leak, we free memory
632 * for signal queue at the time when the state is set.
634 sigqueue_flush(&p->p_sigqueue);
635 sigqueue_flush(&td->td_sigqueue);
638 * We have to wait until after acquiring all locks before
639 * changing p_state. We need to avoid all possible context
640 * switches (including ones from blocking on a mutex) while
641 * marked as a zombie. We also have to set the zombie state
642 * before we release the parent process' proc lock to avoid
643 * a lost wakeup. So, we first call wakeup, then we grab the
644 * sched lock, update the state, and release the parent process'
648 cv_broadcast(&p->p_pwait);
649 sched_exit(p->p_pptr, td);
650 umtx_thread_exit(td);
652 p->p_state = PRS_ZOMBIE;
653 PROC_UNLOCK(p->p_pptr);
656 * Hopefully no one will try to deliver a signal to the process this
659 knlist_destroy(&p->p_klist);
662 * Save our children's rusage information in our exit rusage.
665 ruadd(&p->p_ru, &p->p_rux, &p->p_stats->p_cru, &p->p_crux);
669 * Make sure the scheduler takes this thread out of its tables etc.
670 * This will also release this thread's reference to the ucred.
671 * Other thread parts to release include pcb bits and such.
677 #ifndef _SYS_SYSPROTO_H_
686 sys_abort2(struct thread *td, struct abort2_args *uap)
688 struct proc *p = td->td_proc;
694 * Do it right now so we can log either proper call of abort2(), or
695 * note, that invalid argument was passed. 512 is big enough to
696 * handle 16 arguments' descriptions with additional comments.
698 sb = sbuf_new(NULL, NULL, 512, SBUF_FIXEDLEN);
700 sbuf_printf(sb, "%s(pid %d uid %d) aborted: ",
701 p->p_comm, p->p_pid, td->td_ucred->cr_uid);
703 * Since we can't return from abort2(), send SIGKILL in cases, where
704 * abort2() was called improperly
707 /* Prevent from DoSes from user-space. */
708 if (uap->nargs < 0 || uap->nargs > 16)
710 if (uap->nargs > 0) {
711 if (uap->args == NULL)
713 error = copyin(uap->args, uargs, uap->nargs * sizeof(void *));
718 * Limit size of 'reason' string to 128. Will fit even when
719 * maximal number of arguments was chosen to be logged.
721 if (uap->why != NULL) {
722 error = sbuf_copyin(sb, uap->why, 128);
726 sbuf_printf(sb, "(null)");
728 if (uap->nargs > 0) {
729 sbuf_printf(sb, "(");
730 for (i = 0;i < uap->nargs; i++)
731 sbuf_printf(sb, "%s%p", i == 0 ? "" : ", ", uargs[i]);
732 sbuf_printf(sb, ")");
735 * Final stage: arguments were proper, string has been
736 * successfully copied from userspace, and copying pointers
737 * from user-space succeed.
741 if (sig == SIGKILL) {
743 sbuf_printf(sb, " (Reason text inaccessible)");
747 log(LOG_INFO, "%s", sbuf_data(sb));
756 * The dirty work is handled by kern_wait().
759 owait(struct thread *td, struct owait_args *uap __unused)
763 error = kern_wait(td, WAIT_ANY, &status, 0, NULL);
765 td->td_retval[1] = status;
768 #endif /* COMPAT_43 */
771 * The dirty work is handled by kern_wait().
774 sys_wait4(struct thread *td, struct wait4_args *uap)
776 struct rusage ru, *rup;
779 if (uap->rusage != NULL)
783 error = kern_wait(td, uap->pid, &status, uap->options, rup);
784 if (uap->status != NULL && error == 0)
785 error = copyout(&status, uap->status, sizeof(status));
786 if (uap->rusage != NULL && error == 0)
787 error = copyout(&ru, uap->rusage, sizeof(struct rusage));
792 sys_wait6(struct thread *td, struct wait6_args *uap)
794 struct __wrusage wru, *wrup;
800 idtype = uap->idtype;
803 if (uap->wrusage != NULL)
808 if (uap->info != NULL) {
810 bzero(sip, sizeof(*sip));
815 * We expect all callers of wait6() to know about WEXITED and
818 error = kern_wait6(td, idtype, id, &status, uap->options, wrup, sip);
820 if (uap->status != NULL && error == 0)
821 error = copyout(&status, uap->status, sizeof(status));
822 if (uap->wrusage != NULL && error == 0)
823 error = copyout(&wru, uap->wrusage, sizeof(wru));
824 if (uap->info != NULL && error == 0)
825 error = copyout(&si, uap->info, sizeof(si));
830 * Reap the remains of a zombie process and optionally return status and
831 * rusage. Asserts and will release both the proctree_lock and the process
832 * lock as part of its work.
835 proc_reap(struct thread *td, struct proc *p, int *status, int options)
839 sx_assert(&proctree_lock, SA_XLOCKED);
840 PROC_LOCK_ASSERT(p, MA_OWNED);
841 PROC_SLOCK_ASSERT(p, MA_OWNED);
842 KASSERT(p->p_state == PRS_ZOMBIE, ("proc_reap: !PRS_ZOMBIE"));
848 *status = KW_EXITCODE(p->p_xexit, p->p_xsig);
849 if (options & WNOWAIT) {
851 * Only poll, returning the status. Caller does not wish to
852 * release the proc struct just yet.
855 sx_xunlock(&proctree_lock);
860 sigqueue_take(p->p_ksi);
864 * If we got the child via a ptrace 'attach', we need to give it back
867 if (p->p_oppid != 0 && p->p_oppid != p->p_pptr->p_pid) {
869 t = proc_realparent(p);
873 "wait: traced child %d moved back to parent %d", p->p_pid,
878 pksignal(t, SIGCHLD, p->p_ksi);
880 cv_broadcast(&p->p_pwait);
882 sx_xunlock(&proctree_lock);
889 * Remove other references to this process to ensure we have an
890 * exclusive reference.
892 sx_xlock(&allproc_lock);
893 LIST_REMOVE(p, p_list); /* off zombproc */
894 sx_xunlock(&allproc_lock);
895 LIST_REMOVE(p, p_sibling);
896 reaper_abandon_children(p, true);
897 LIST_REMOVE(p, p_reapsibling);
902 if (p->p_procdesc != NULL)
904 sx_xunlock(&proctree_lock);
907 * Removal from allproc list and process group list paired with
908 * PROC_LOCK which was executed during that time should guarantee
909 * nothing can reach this process anymore. As such further locking
912 p->p_xexit = p->p_xsig = 0; /* XXX: why? */
915 ruadd(&q->p_stats->p_cru, &q->p_crux, &p->p_ru, &p->p_rux);
919 * Decrement the count of procs running with this uid.
921 (void)chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0);
924 * Destroy resource accounting information associated with the process.
929 racct_sub(p, RACCT_NPROC, 1);
936 * Free credentials, arguments, and sigacts.
939 proc_set_cred(p, NULL);
940 pargs_drop(p->p_args);
942 sigacts_free(p->p_sigacts);
946 * Do any thread-system specific cleanups.
951 * Give vm and machine-dependent layer a chance to free anything that
952 * cpu_exit couldn't release while still running in process context.
959 * Free any domain policy that's still hiding around.
961 vm_domain_policy_cleanup(&p->p_vm_dom_policy);
963 KASSERT(FIRST_THREAD_IN_PROC(p),
964 ("proc_reap: no residual thread!"));
965 uma_zfree(proc_zone, p);
966 sx_xlock(&allproc_lock);
968 sx_xunlock(&allproc_lock);
972 proc_to_reap(struct thread *td, struct proc *p, idtype_t idtype, id_t id,
973 int *status, int options, struct __wrusage *wrusage, siginfo_t *siginfo,
978 sx_assert(&proctree_lock, SA_XLOCKED);
984 if (p->p_procdesc != NULL) {
990 if (p->p_pid != (pid_t)id) {
996 if (p->p_pgid != (pid_t)id) {
1002 if (p->p_session->s_sid != (pid_t)id) {
1008 if (p->p_ucred->cr_uid != (uid_t)id) {
1014 if (p->p_ucred->cr_gid != (gid_t)id) {
1020 if (p->p_ucred->cr_prison->pr_id != (int)id) {
1026 * It seems that the thread structures get zeroed out
1027 * at process exit. This makes it impossible to
1028 * support P_SETID, P_CID or P_CPUID.
1035 if (p_canwait(td, p)) {
1040 if (((options & WEXITED) == 0) && (p->p_state == PRS_ZOMBIE)) {
1046 * This special case handles a kthread spawned by linux_clone
1047 * (see linux_misc.c). The linux_wait4 and linux_waitpid
1048 * functions need to be able to distinguish between waiting
1049 * on a process and waiting on a thread. It is a thread if
1050 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option
1051 * signifies we want to wait for threads and not processes.
1053 if ((p->p_sigparent != SIGCHLD) ^
1054 ((options & WLINUXCLONE) != 0)) {
1059 if (siginfo != NULL) {
1060 bzero(siginfo, sizeof(*siginfo));
1061 siginfo->si_errno = 0;
1064 * SUSv4 requires that the si_signo value is always
1065 * SIGCHLD. Obey it despite the rfork(2) interface
1066 * allows to request other signal for child exit
1069 siginfo->si_signo = SIGCHLD;
1072 * This is still a rough estimate. We will fix the
1073 * cases TRAPPED, STOPPED, and CONTINUED later.
1075 if (WCOREDUMP(p->p_xsig)) {
1076 siginfo->si_code = CLD_DUMPED;
1077 siginfo->si_status = WTERMSIG(p->p_xsig);
1078 } else if (WIFSIGNALED(p->p_xsig)) {
1079 siginfo->si_code = CLD_KILLED;
1080 siginfo->si_status = WTERMSIG(p->p_xsig);
1082 siginfo->si_code = CLD_EXITED;
1083 siginfo->si_status = p->p_xexit;
1086 siginfo->si_pid = p->p_pid;
1087 siginfo->si_uid = p->p_ucred->cr_uid;
1090 * The si_addr field would be useful additional
1091 * detail, but apparently the PC value may be lost
1092 * when we reach this point. bzero() above sets
1093 * siginfo->si_addr to NULL.
1098 * There should be no reason to limit resources usage info to
1099 * exited processes only. A snapshot about any resources used
1100 * by a stopped process may be exactly what is needed.
1102 if (wrusage != NULL) {
1103 rup = &wrusage->wru_self;
1106 calcru(p, &rup->ru_utime, &rup->ru_stime);
1109 rup = &wrusage->wru_children;
1110 *rup = p->p_stats->p_cru;
1111 calccru(p, &rup->ru_utime, &rup->ru_stime);
1114 if (p->p_state == PRS_ZOMBIE && !check_only) {
1116 proc_reap(td, p, status, options);
1124 kern_wait(struct thread *td, pid_t pid, int *status, int options,
1125 struct rusage *rusage)
1127 struct __wrusage wru, *wrup;
1133 * Translate the special pid values into the (idtype, pid)
1134 * pair for kern_wait6. The WAIT_MYPGRP case is handled by
1135 * kern_wait6() on its own.
1137 if (pid == WAIT_ANY) {
1140 } else if (pid < 0) {
1154 * For backward compatibility we implicitly add flags WEXITED
1155 * and WTRAPPED here.
1157 options |= WEXITED | WTRAPPED;
1158 ret = kern_wait6(td, idtype, id, status, options, wrup, NULL);
1160 *rusage = wru.wru_self;
1165 kern_wait6(struct thread *td, idtype_t idtype, id_t id, int *status,
1166 int options, struct __wrusage *wrusage, siginfo_t *siginfo)
1170 int error, nfound, ret;
1172 AUDIT_ARG_VALUE((int)idtype); /* XXX - This is likely wrong! */
1173 AUDIT_ARG_PID((pid_t)id); /* XXX - This may be wrong! */
1174 AUDIT_ARG_VALUE(options);
1178 if ((pid_t)id == WAIT_MYPGRP && (idtype == P_PID || idtype == P_PGID)) {
1180 id = (id_t)q->p_pgid;
1185 /* If we don't know the option, just return. */
1186 if ((options & ~(WUNTRACED | WNOHANG | WCONTINUED | WNOWAIT |
1187 WEXITED | WTRAPPED | WLINUXCLONE)) != 0)
1189 if ((options & (WEXITED | WUNTRACED | WCONTINUED | WTRAPPED)) == 0) {
1191 * We will be unable to find any matching processes,
1192 * because there are no known events to look for.
1193 * Prefer to return error instead of blocking
1200 if (q->p_flag & P_STATCHILD) {
1202 q->p_flag &= ~P_STATCHILD;
1206 sx_xlock(&proctree_lock);
1207 LIST_FOREACH(p, &q->p_children, p_sibling) {
1209 ret = proc_to_reap(td, p, idtype, id, status, options,
1210 wrusage, siginfo, 0);
1216 td->td_retval[0] = pid;
1223 if ((options & WTRAPPED) != 0 &&
1224 (p->p_flag & P_TRACED) != 0 &&
1225 (p->p_flag & (P_STOPPED_TRACE | P_STOPPED_SIG)) != 0 &&
1226 (p->p_suspcount == p->p_numthreads) &&
1227 ((p->p_flag & P_WAITED) == 0)) {
1229 if ((options & WNOWAIT) == 0)
1230 p->p_flag |= P_WAITED;
1231 sx_xunlock(&proctree_lock);
1234 *status = W_STOPCODE(p->p_xsig);
1235 if (siginfo != NULL) {
1236 siginfo->si_status = p->p_xsig;
1237 siginfo->si_code = CLD_TRAPPED;
1239 if ((options & WNOWAIT) == 0) {
1241 sigqueue_take(p->p_ksi);
1246 "wait: returning trapped pid %d status %#x (xstat %d) xthread %d",
1247 p->p_pid, W_STOPCODE(p->p_xsig), p->p_xsig,
1248 p->p_xthread != NULL ? p->p_xthread->td_tid : -1);
1250 td->td_retval[0] = pid;
1253 if ((options & WUNTRACED) != 0 &&
1254 (p->p_flag & P_STOPPED_SIG) != 0 &&
1255 (p->p_suspcount == p->p_numthreads) &&
1256 ((p->p_flag & P_WAITED) == 0)) {
1258 if ((options & WNOWAIT) == 0)
1259 p->p_flag |= P_WAITED;
1260 sx_xunlock(&proctree_lock);
1263 *status = W_STOPCODE(p->p_xsig);
1264 if (siginfo != NULL) {
1265 siginfo->si_status = p->p_xsig;
1266 siginfo->si_code = CLD_STOPPED;
1268 if ((options & WNOWAIT) == 0) {
1270 sigqueue_take(p->p_ksi);
1275 td->td_retval[0] = pid;
1279 if ((options & WCONTINUED) != 0 &&
1280 (p->p_flag & P_CONTINUED) != 0) {
1281 sx_xunlock(&proctree_lock);
1282 if ((options & WNOWAIT) == 0) {
1283 p->p_flag &= ~P_CONTINUED;
1285 sigqueue_take(p->p_ksi);
1292 if (siginfo != NULL) {
1293 siginfo->si_status = SIGCONT;
1294 siginfo->si_code = CLD_CONTINUED;
1296 td->td_retval[0] = pid;
1303 * Look in the orphans list too, to allow the parent to
1304 * collect it's child exit status even if child is being
1307 * Debugger detaches from the parent upon successful
1308 * switch-over from parent to child. At this point due to
1309 * re-parenting the parent loses the child to debugger and a
1310 * wait4(2) call would report that it has no children to wait
1311 * for. By maintaining a list of orphans we allow the parent
1312 * to successfully wait until the child becomes a zombie.
1315 LIST_FOREACH(p, &q->p_orphans, p_orphan) {
1316 ret = proc_to_reap(td, p, idtype, id, NULL, options,
1319 KASSERT(ret != -1, ("reaped an orphan (pid %d)",
1320 (int)td->td_retval[0]));
1327 sx_xunlock(&proctree_lock);
1330 if (options & WNOHANG) {
1331 sx_xunlock(&proctree_lock);
1332 td->td_retval[0] = 0;
1336 sx_xunlock(&proctree_lock);
1337 if (q->p_flag & P_STATCHILD) {
1338 q->p_flag &= ~P_STATCHILD;
1341 error = msleep(q, &q->p_mtx, PWAIT | PCATCH, "wait", 0);
1349 * Make process 'parent' the new parent of process 'child'.
1350 * Must be called with an exclusive hold of proctree lock.
1353 proc_reparent(struct proc *child, struct proc *parent)
1356 sx_assert(&proctree_lock, SX_XLOCKED);
1357 PROC_LOCK_ASSERT(child, MA_OWNED);
1358 if (child->p_pptr == parent)
1361 PROC_LOCK(child->p_pptr);
1362 sigqueue_take(child->p_ksi);
1363 PROC_UNLOCK(child->p_pptr);
1364 LIST_REMOVE(child, p_sibling);
1365 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
1367 clear_orphan(child);
1368 if (child->p_flag & P_TRACED) {
1369 if (LIST_EMPTY(&child->p_pptr->p_orphans)) {
1370 child->p_treeflag |= P_TREE_FIRST_ORPHAN;
1371 LIST_INSERT_HEAD(&child->p_pptr->p_orphans, child,
1374 LIST_INSERT_AFTER(LIST_FIRST(&child->p_pptr->p_orphans),
1377 child->p_treeflag |= P_TREE_ORPHANED;
1380 child->p_pptr = parent;