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>
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");
100 proc_realparent(struct proc *child)
102 struct proc *p, *parent;
104 sx_assert(&proctree_lock, SX_LOCKED);
105 if ((child->p_treeflag & P_TREE_ORPHANED) == 0)
106 return (child->p_pptr->p_pid == child->p_oppid ?
107 child->p_pptr : initproc);
108 for (p = child; (p->p_treeflag & P_TREE_FIRST_ORPHAN) == 0;) {
109 /* Cannot use LIST_PREV(), since the list head is not known. */
110 p = __containerof(p->p_orphan.le_prev, struct proc,
112 KASSERT((p->p_treeflag & P_TREE_ORPHANED) != 0,
113 ("missing P_ORPHAN %p", p));
115 parent = __containerof(p->p_orphan.le_prev, struct proc,
121 reaper_abandon_children(struct proc *p, bool exiting)
123 struct proc *p1, *p2, *ptmp;
125 sx_assert(&proctree_lock, SX_LOCKED);
126 KASSERT(p != initproc, ("reaper_abandon_children for initproc"));
127 if ((p->p_treeflag & P_TREE_REAPER) == 0)
130 LIST_FOREACH_SAFE(p2, &p->p_reaplist, p_reapsibling, ptmp) {
131 LIST_REMOVE(p2, p_reapsibling);
133 p2->p_reapsubtree = p->p_reapsubtree;
134 LIST_INSERT_HEAD(&p1->p_reaplist, p2, p_reapsibling);
135 if (exiting && p2->p_pptr == p) {
137 proc_reparent(p2, p1, true);
141 KASSERT(LIST_EMPTY(&p->p_reaplist), ("p_reaplist not empty"));
142 p->p_treeflag &= ~P_TREE_REAPER;
146 reaper_clear(struct proc *p)
151 sx_assert(&proctree_lock, SX_LOCKED);
152 LIST_REMOVE(p, p_reapsibling);
153 if (p->p_reapsubtree == 1)
156 LIST_FOREACH(p1, &p->p_reaper->p_reaplist, p_reapsibling) {
157 if (p1->p_reapsubtree == p->p_reapsubtree) {
163 proc_id_clear(PROC_ID_REAP, p->p_reapsubtree);
167 proc_clear_orphan(struct proc *p)
171 sx_assert(&proctree_lock, SA_XLOCKED);
172 if ((p->p_treeflag & P_TREE_ORPHANED) == 0)
174 if ((p->p_treeflag & P_TREE_FIRST_ORPHAN) != 0) {
175 p1 = LIST_NEXT(p, p_orphan);
177 p1->p_treeflag |= P_TREE_FIRST_ORPHAN;
178 p->p_treeflag &= ~P_TREE_FIRST_ORPHAN;
180 LIST_REMOVE(p, p_orphan);
181 p->p_treeflag &= ~P_TREE_ORPHANED;
185 * exit -- death of process.
188 sys_sys_exit(struct thread *td, struct sys_exit_args *uap)
191 exit1(td, uap->rval, 0);
196 * Exit: deallocate address space and other resources, change proc state to
197 * zombie, and unlink proc from allproc and parent's lists. Save exit status
198 * and rusage for wait(). Check for child processes and orphan them.
201 exit1(struct thread *td, int rval, int signo)
203 struct proc *p, *nq, *q, *t;
205 ksiginfo_t *ksi, *ksi1;
208 mtx_assert(&Giant, MA_NOTOWNED);
209 KASSERT(rval == 0 || signo == 0, ("exit1 rv %d sig %d", rval, signo));
213 * XXX in case we're rebooting we just let init die in order to
214 * work around an unsolved stack overflow seen very late during
215 * shutdown on sparc64 when the gmirror worker process exists.
216 * XXX what to do now that sparc64 is gone... remove if?
218 if (p == initproc && rebooting == 0) {
219 printf("init died (signal %d, exit %d)\n", signo, rval);
220 panic("Going nowhere without my init!");
224 * Deref SU mp, since the thread does not return to userspace.
226 td_softdep_cleanup(td);
229 * MUST abort all other threads before proceeding past here.
233 * First check if some other thread or external request got
234 * here before us. If so, act appropriately: exit or suspend.
235 * We must ensure that stop requests are handled before we set
238 thread_suspend_check(0);
239 while (p->p_flag & P_HADTHREADS) {
241 * Kill off the other threads. This requires
242 * some co-operation from other parts of the kernel
243 * so it may not be instantaneous. With this state set
244 * any thread entering the kernel from userspace will
245 * thread_exit() in trap(). Any thread attempting to
246 * sleep will return immediately with EINTR or EWOULDBLOCK
247 * which will hopefully force them to back out to userland
248 * freeing resources as they go. Any thread attempting
249 * to return to userland will thread_exit() from userret().
250 * thread_exit() will unsuspend us when the last of the
251 * other threads exits.
252 * If there is already a thread singler after resumption,
253 * calling thread_single will fail; in that case, we just
254 * re-check all suspension request, the thread should
255 * either be suspended there or exit.
257 if (!thread_single(p, SINGLE_EXIT))
259 * All other activity in this process is now
260 * stopped. Threading support has been turned
265 * Recheck for new stop or suspend requests which
266 * might appear while process lock was dropped in
269 thread_suspend_check(0);
271 KASSERT(p->p_numthreads == 1,
272 ("exit1: proc %p exiting with %d threads", p, p->p_numthreads));
273 racct_sub(p, RACCT_NTHR, 1);
275 /* Let event handler change exit status */
280 * Ignore any pending request to stop due to a stop signal.
281 * Once P_WEXIT is set, future requests will be ignored as
284 p->p_flag &= ~P_STOPPED_SIG;
285 KASSERT(!P_SHOULDSTOP(p), ("exiting process is stopped"));
287 /* Note that we are exiting. */
288 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_DIRECT_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.
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_safe(&p->p_itcallout, CS_EXECUTING, NULL) == 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 umtx_thread_exit(td);
360 * Reset any sigio structures pointing to us as a result of
361 * F_SETOWN with our pid.
363 funsetownlst(&p->p_sigiolst);
366 * Close open files and release open-file table.
372 * If this thread tickled GEOM, we need to wait for the giggling to
373 * stop before we return to userland
375 if (td->td_pflags & TDP_GEOM)
379 * Remove ourself from our leader's peer list and wake our leader.
381 if (p->p_leader->p_peers != NULL) {
382 mtx_lock(&ppeers_lock);
383 if (p->p_leader->p_peers != NULL) {
385 while (q->p_peers != p)
387 q->p_peers = p->p_peers;
390 mtx_unlock(&ppeers_lock);
395 (void)acct_process(td);
401 * Release reference to text vnode
403 if (p->p_textvp != NULL) {
409 * Release our limits structure.
411 lim_free(p->p_limit);
417 * Call machine-dependent code to release any
418 * machine-dependent resources other than the address space.
419 * The address space is released by "vmspace_exitfree(p)" in
424 WITNESS_WARN(WARN_PANIC, NULL, "process (pid %d) exiting", p->p_pid);
427 * Remove from allproc. It still sits in the hash.
429 sx_xlock(&allproc_lock);
430 LIST_REMOVE(p, p_list);
431 sx_xunlock(&allproc_lock);
433 sx_xlock(&proctree_lock);
435 p->p_flag &= ~(P_TRACED | P_PPWAIT | P_PPTRACE);
439 * Reparent all children processes:
440 * - traced ones to the original parent (or init if we are that parent)
443 q = LIST_FIRST(&p->p_children);
444 if (q != NULL) /* only need this if any child is S_ZOMB */
446 for (; q != NULL; q = nq) {
447 nq = LIST_NEXT(q, p_sibling);
448 ksi = ksiginfo_alloc(TRUE);
450 q->p_sigparent = SIGCHLD;
452 if ((q->p_flag & P_TRACED) == 0) {
453 proc_reparent(q, q->p_reaper, true);
454 if (q->p_state == PRS_ZOMBIE) {
456 * Inform reaper about the reparented
457 * zombie, since wait(2) has something
458 * new to report. Guarantee queueing
459 * of the SIGCHLD signal, similar to
460 * the _exit() behaviour, by providing
461 * our ksiginfo. Ksi is freed by the
464 if (q->p_ksi == NULL) {
467 ksiginfo_copy(q->p_ksi, ksi);
468 ksi->ksi_flags |= KSI_INS;
472 PROC_LOCK(q->p_reaper);
473 pksignal(q->p_reaper, SIGCHLD, ksi1);
474 PROC_UNLOCK(q->p_reaper);
475 } else if (q->p_pdeathsig > 0) {
477 * The child asked to received a signal
480 kern_psignal(q, q->p_pdeathsig);
484 * Traced processes are killed since their existence
485 * means someone is screwing up.
487 t = proc_realparent(q);
489 proc_reparent(q, q->p_reaper, true);
492 proc_reparent(q, t, true);
496 * Since q was found on our children list, the
497 * proc_reparent() call moved q to the orphan
498 * list due to present P_TRACED flag. Clear
499 * orphan link for q now while q is locked.
501 proc_clear_orphan(q);
502 q->p_flag &= ~(P_TRACED | P_STOPPED_TRACE);
503 q->p_flag2 &= ~P2_PTRACE_FSTP;
505 FOREACH_THREAD_IN_PROC(q, tdt) {
506 tdt->td_dbgflags &= ~(TDB_SUSPEND | TDB_XSIG |
509 kern_psignal(q, SIGKILL);
517 * Also get rid of our orphans.
519 while ((q = LIST_FIRST(&p->p_orphans)) != NULL) {
521 KASSERT(q->p_oppid == p->p_pid,
522 ("orphan %p of %p has unexpected oppid %d", q, p,
524 q->p_oppid = q->p_reaper->p_pid;
527 * If we are the real parent of this process
528 * but it has been reparented to a debugger, then
529 * check if it asked for a signal when we exit.
531 if (q->p_pdeathsig > 0)
532 kern_psignal(q, q->p_pdeathsig);
533 CTR2(KTR_PTRACE, "exit: pid %d, clearing orphan %d", p->p_pid,
535 proc_clear_orphan(q);
540 if (SDT_PROBES_ENABLED()) {
541 int reason = CLD_EXITED;
542 if (WCOREDUMP(signo))
544 else if (WIFSIGNALED(signo))
546 SDT_PROBE1(proc, , , exit, reason);
550 /* Save exit status. */
556 * Tell the DTrace fasttrap provider about the exit if it
557 * has declared an interest.
559 if (dtrace_fasttrap_exit)
560 dtrace_fasttrap_exit(p);
564 * Notify interested parties of our demise.
566 KNOTE_LOCKED(p->p_klist, NOTE_EXIT);
569 * If this is a process with a descriptor, we may not need to deliver
570 * a signal to the parent. proctree_lock is held over
571 * procdesc_exit() to serialize concurrent calls to close() and
575 if (p->p_procdesc == NULL || procdesc_exit(p)) {
577 * Notify parent that we're gone. If parent has the
578 * PS_NOCLDWAIT flag set, or if the handler is set to SIG_IGN,
579 * notify process 1 instead (and hope it will handle this
582 PROC_LOCK(p->p_pptr);
583 mtx_lock(&p->p_pptr->p_sigacts->ps_mtx);
584 if (p->p_pptr->p_sigacts->ps_flag &
585 (PS_NOCLDWAIT | PS_CLDSIGIGN)) {
588 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
591 proc_reparent(p, p->p_reaper, true);
592 p->p_sigparent = SIGCHLD;
593 PROC_LOCK(p->p_pptr);
596 * Notify parent, so in case he was wait(2)ing or
597 * executing waitpid(2) with our pid, he will
602 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
604 if (p->p_pptr == p->p_reaper || p->p_pptr == initproc) {
606 } else if (p->p_sigparent != 0) {
607 if (p->p_sigparent == SIGCHLD) {
609 } else { /* LINUX thread */
614 PROC_LOCK(p->p_pptr);
615 sx_xunlock(&proctree_lock);
617 if (signal_parent == 1) {
619 } else if (signal_parent == 2) {
620 kern_psignal(p->p_pptr, p->p_sigparent);
623 /* Tell the prison that we are gone. */
624 prison_proc_free(p->p_ucred->cr_prison);
627 * The state PRS_ZOMBIE prevents other proesses from sending
628 * signal to the process, to avoid memory leak, we free memory
629 * for signal queue at the time when the state is set.
631 sigqueue_flush(&p->p_sigqueue);
632 sigqueue_flush(&td->td_sigqueue);
635 * We have to wait until after acquiring all locks before
636 * changing p_state. We need to avoid all possible context
637 * switches (including ones from blocking on a mutex) while
638 * marked as a zombie. We also have to set the zombie state
639 * before we release the parent process' proc lock to avoid
640 * a lost wakeup. So, we first call wakeup, then we grab the
641 * sched lock, update the state, and release the parent process'
645 cv_broadcast(&p->p_pwait);
646 sched_exit(p->p_pptr, td);
648 p->p_state = PRS_ZOMBIE;
649 PROC_UNLOCK(p->p_pptr);
652 * Save our children's rusage information in our exit rusage.
655 ruadd(&p->p_ru, &p->p_rux, &p->p_stats->p_cru, &p->p_crux);
659 * Make sure the scheduler takes this thread out of its tables etc.
660 * This will also release this thread's reference to the ucred.
661 * Other thread parts to release include pcb bits and such.
666 #ifndef _SYS_SYSPROTO_H_
675 sys_abort2(struct thread *td, struct abort2_args *uap)
677 struct proc *p = td->td_proc;
683 * Do it right now so we can log either proper call of abort2(), or
684 * note, that invalid argument was passed. 512 is big enough to
685 * handle 16 arguments' descriptions with additional comments.
687 sb = sbuf_new(NULL, NULL, 512, SBUF_FIXEDLEN);
689 sbuf_printf(sb, "%s(pid %d uid %d) aborted: ",
690 p->p_comm, p->p_pid, td->td_ucred->cr_uid);
692 * Since we can't return from abort2(), send SIGKILL in cases, where
693 * abort2() was called improperly
696 /* Prevent from DoSes from user-space. */
697 if (uap->nargs < 0 || uap->nargs > 16)
699 if (uap->nargs > 0) {
700 if (uap->args == NULL)
702 error = copyin(uap->args, uargs, uap->nargs * sizeof(void *));
707 * Limit size of 'reason' string to 128. Will fit even when
708 * maximal number of arguments was chosen to be logged.
710 if (uap->why != NULL) {
711 error = sbuf_copyin(sb, uap->why, 128);
715 sbuf_printf(sb, "(null)");
717 if (uap->nargs > 0) {
718 sbuf_printf(sb, "(");
719 for (i = 0;i < uap->nargs; i++)
720 sbuf_printf(sb, "%s%p", i == 0 ? "" : ", ", uargs[i]);
721 sbuf_printf(sb, ")");
724 * Final stage: arguments were proper, string has been
725 * successfully copied from userspace, and copying pointers
726 * from user-space succeed.
730 if (sig == SIGKILL) {
732 sbuf_printf(sb, " (Reason text inaccessible)");
736 log(LOG_INFO, "%s", sbuf_data(sb));
744 * The dirty work is handled by kern_wait().
747 owait(struct thread *td, struct owait_args *uap __unused)
751 error = kern_wait(td, WAIT_ANY, &status, 0, NULL);
753 td->td_retval[1] = status;
756 #endif /* COMPAT_43 */
759 * The dirty work is handled by kern_wait().
762 sys_wait4(struct thread *td, struct wait4_args *uap)
764 struct rusage ru, *rup;
767 if (uap->rusage != NULL)
771 error = kern_wait(td, uap->pid, &status, uap->options, rup);
772 if (uap->status != NULL && error == 0 && td->td_retval[0] != 0)
773 error = copyout(&status, uap->status, sizeof(status));
774 if (uap->rusage != NULL && error == 0 && td->td_retval[0] != 0)
775 error = copyout(&ru, uap->rusage, sizeof(struct rusage));
780 sys_wait6(struct thread *td, struct wait6_args *uap)
782 struct __wrusage wru, *wrup;
788 idtype = uap->idtype;
791 if (uap->wrusage != NULL)
796 if (uap->info != NULL) {
798 bzero(sip, sizeof(*sip));
803 * We expect all callers of wait6() to know about WEXITED and
806 error = kern_wait6(td, idtype, id, &status, uap->options, wrup, sip);
808 if (uap->status != NULL && error == 0 && td->td_retval[0] != 0)
809 error = copyout(&status, uap->status, sizeof(status));
810 if (uap->wrusage != NULL && error == 0 && td->td_retval[0] != 0)
811 error = copyout(&wru, uap->wrusage, sizeof(wru));
812 if (uap->info != NULL && error == 0)
813 error = copyout(&si, uap->info, sizeof(si));
818 * Reap the remains of a zombie process and optionally return status and
819 * rusage. Asserts and will release both the proctree_lock and the process
820 * lock as part of its work.
823 proc_reap(struct thread *td, struct proc *p, int *status, int options)
827 sx_assert(&proctree_lock, SA_XLOCKED);
828 PROC_LOCK_ASSERT(p, MA_OWNED);
829 KASSERT(p->p_state == PRS_ZOMBIE, ("proc_reap: !PRS_ZOMBIE"));
831 mtx_spin_wait_unlocked(&p->p_slock);
836 *status = KW_EXITCODE(p->p_xexit, p->p_xsig);
837 if (options & WNOWAIT) {
839 * Only poll, returning the status. Caller does not wish to
840 * release the proc struct just yet.
843 sx_xunlock(&proctree_lock);
848 sigqueue_take(p->p_ksi);
852 * If we got the child via a ptrace 'attach', we need to give it back
855 if (p->p_oppid != p->p_pptr->p_pid) {
857 t = proc_realparent(p);
861 "wait: traced child %d moved back to parent %d", p->p_pid,
863 proc_reparent(p, t, false);
865 pksignal(t, SIGCHLD, p->p_ksi);
867 cv_broadcast(&p->p_pwait);
869 sx_xunlock(&proctree_lock);
875 * Remove other references to this process to ensure we have an
876 * exclusive reference.
878 sx_xlock(PIDHASHLOCK(p->p_pid));
879 LIST_REMOVE(p, p_hash);
880 sx_xunlock(PIDHASHLOCK(p->p_pid));
881 LIST_REMOVE(p, p_sibling);
882 reaper_abandon_children(p, true);
885 proc_clear_orphan(p);
888 if (p->p_procdesc != NULL)
890 sx_xunlock(&proctree_lock);
892 proc_id_clear(PROC_ID_PID, p->p_pid);
895 knlist_detach(p->p_klist);
900 * Removal from allproc list and process group list paired with
901 * PROC_LOCK which was executed during that time should guarantee
902 * nothing can reach this process anymore. As such further locking
905 p->p_xexit = p->p_xsig = 0; /* XXX: why? */
908 ruadd(&q->p_stats->p_cru, &q->p_crux, &p->p_ru, &p->p_rux);
912 * Decrement the count of procs running with this uid.
914 (void)chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0);
917 * Destroy resource accounting information associated with the process.
922 racct_sub(p, RACCT_NPROC, 1);
929 * Free credentials, arguments, and sigacts.
932 proc_set_cred(p, NULL);
933 pargs_drop(p->p_args);
935 sigacts_free(p->p_sigacts);
939 * Do any thread-system specific cleanups.
944 * Give vm and machine-dependent layer a chance to free anything that
945 * cpu_exit couldn't release while still running in process context.
952 KASSERT(FIRST_THREAD_IN_PROC(p),
953 ("proc_reap: no residual thread!"));
954 uma_zfree(proc_zone, p);
955 atomic_add_int(&nprocs, -1);
959 proc_to_reap(struct thread *td, struct proc *p, idtype_t idtype, id_t id,
960 int *status, int options, struct __wrusage *wrusage, siginfo_t *siginfo,
965 sx_assert(&proctree_lock, SA_XLOCKED);
971 if (p->p_procdesc == NULL ||
972 (p->p_pptr == td->td_proc &&
973 (p->p_flag & P_TRACED) != 0)) {
980 if (p->p_pid != (pid_t)id) {
986 if (p->p_pgid != (pid_t)id) {
992 if (p->p_session->s_sid != (pid_t)id) {
998 if (p->p_ucred->cr_uid != (uid_t)id) {
1004 if (p->p_ucred->cr_gid != (gid_t)id) {
1010 if (p->p_ucred->cr_prison->pr_id != (int)id) {
1016 * It seems that the thread structures get zeroed out
1017 * at process exit. This makes it impossible to
1018 * support P_SETID, P_CID or P_CPUID.
1025 if (p_canwait(td, p)) {
1030 if (((options & WEXITED) == 0) && (p->p_state == PRS_ZOMBIE)) {
1036 * This special case handles a kthread spawned by linux_clone
1037 * (see linux_misc.c). The linux_wait4 and linux_waitpid
1038 * functions need to be able to distinguish between waiting
1039 * on a process and waiting on a thread. It is a thread if
1040 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option
1041 * signifies we want to wait for threads and not processes.
1043 if ((p->p_sigparent != SIGCHLD) ^
1044 ((options & WLINUXCLONE) != 0)) {
1049 if (siginfo != NULL) {
1050 bzero(siginfo, sizeof(*siginfo));
1051 siginfo->si_errno = 0;
1054 * SUSv4 requires that the si_signo value is always
1055 * SIGCHLD. Obey it despite the rfork(2) interface
1056 * allows to request other signal for child exit
1059 siginfo->si_signo = SIGCHLD;
1062 * This is still a rough estimate. We will fix the
1063 * cases TRAPPED, STOPPED, and CONTINUED later.
1065 if (WCOREDUMP(p->p_xsig)) {
1066 siginfo->si_code = CLD_DUMPED;
1067 siginfo->si_status = WTERMSIG(p->p_xsig);
1068 } else if (WIFSIGNALED(p->p_xsig)) {
1069 siginfo->si_code = CLD_KILLED;
1070 siginfo->si_status = WTERMSIG(p->p_xsig);
1072 siginfo->si_code = CLD_EXITED;
1073 siginfo->si_status = p->p_xexit;
1076 siginfo->si_pid = p->p_pid;
1077 siginfo->si_uid = p->p_ucred->cr_uid;
1080 * The si_addr field would be useful additional
1081 * detail, but apparently the PC value may be lost
1082 * when we reach this point. bzero() above sets
1083 * siginfo->si_addr to NULL.
1088 * There should be no reason to limit resources usage info to
1089 * exited processes only. A snapshot about any resources used
1090 * by a stopped process may be exactly what is needed.
1092 if (wrusage != NULL) {
1093 rup = &wrusage->wru_self;
1096 calcru(p, &rup->ru_utime, &rup->ru_stime);
1099 rup = &wrusage->wru_children;
1100 *rup = p->p_stats->p_cru;
1101 calccru(p, &rup->ru_utime, &rup->ru_stime);
1104 if (p->p_state == PRS_ZOMBIE && !check_only) {
1105 proc_reap(td, p, status, options);
1112 kern_wait(struct thread *td, pid_t pid, int *status, int options,
1113 struct rusage *rusage)
1115 struct __wrusage wru, *wrup;
1121 * Translate the special pid values into the (idtype, pid)
1122 * pair for kern_wait6. The WAIT_MYPGRP case is handled by
1123 * kern_wait6() on its own.
1125 if (pid == WAIT_ANY) {
1128 } else if (pid < 0) {
1142 * For backward compatibility we implicitly add flags WEXITED
1143 * and WTRAPPED here.
1145 options |= WEXITED | WTRAPPED;
1146 ret = kern_wait6(td, idtype, id, status, options, wrup, NULL);
1148 *rusage = wru.wru_self;
1153 report_alive_proc(struct thread *td, struct proc *p, siginfo_t *siginfo,
1154 int *status, int options, int si_code)
1158 PROC_LOCK_ASSERT(p, MA_OWNED);
1159 sx_assert(&proctree_lock, SA_XLOCKED);
1160 MPASS(si_code == CLD_TRAPPED || si_code == CLD_STOPPED ||
1161 si_code == CLD_CONTINUED);
1163 cont = si_code == CLD_CONTINUED;
1164 if ((options & WNOWAIT) == 0) {
1166 p->p_flag &= ~P_CONTINUED;
1168 p->p_flag |= P_WAITED;
1169 PROC_LOCK(td->td_proc);
1170 sigqueue_take(p->p_ksi);
1171 PROC_UNLOCK(td->td_proc);
1173 sx_xunlock(&proctree_lock);
1174 if (siginfo != NULL) {
1175 siginfo->si_code = si_code;
1176 siginfo->si_status = cont ? SIGCONT : p->p_xsig;
1179 *status = cont ? SIGCONT : W_STOPCODE(p->p_xsig);
1181 td->td_retval[0] = p->p_pid;
1185 kern_wait6(struct thread *td, idtype_t idtype, id_t id, int *status,
1186 int options, struct __wrusage *wrusage, siginfo_t *siginfo)
1190 int error, nfound, ret;
1193 AUDIT_ARG_VALUE((int)idtype); /* XXX - This is likely wrong! */
1194 AUDIT_ARG_PID((pid_t)id); /* XXX - This may be wrong! */
1195 AUDIT_ARG_VALUE(options);
1199 if ((pid_t)id == WAIT_MYPGRP && (idtype == P_PID || idtype == P_PGID)) {
1201 id = (id_t)q->p_pgid;
1206 /* If we don't know the option, just return. */
1207 if ((options & ~(WUNTRACED | WNOHANG | WCONTINUED | WNOWAIT |
1208 WEXITED | WTRAPPED | WLINUXCLONE)) != 0)
1210 if ((options & (WEXITED | WUNTRACED | WCONTINUED | WTRAPPED)) == 0) {
1212 * We will be unable to find any matching processes,
1213 * because there are no known events to look for.
1214 * Prefer to return error instead of blocking
1221 if (q->p_flag & P_STATCHILD) {
1223 q->p_flag &= ~P_STATCHILD;
1226 sx_xlock(&proctree_lock);
1229 LIST_FOREACH(p, &q->p_children, p_sibling) {
1231 ret = proc_to_reap(td, p, idtype, id, status, options,
1232 wrusage, siginfo, 0);
1235 else if (ret != 1) {
1236 td->td_retval[0] = pid;
1241 PROC_LOCK_ASSERT(p, MA_OWNED);
1243 if ((options & WTRAPPED) != 0 &&
1244 (p->p_flag & P_TRACED) != 0) {
1247 ((p->p_flag & (P_STOPPED_TRACE | P_STOPPED_SIG)) &&
1248 p->p_suspcount == p->p_numthreads &&
1249 (p->p_flag & P_WAITED) == 0);
1253 "wait: returning trapped pid %d status %#x "
1254 "(xstat %d) xthread %d",
1255 p->p_pid, W_STOPCODE(p->p_xsig), p->p_xsig,
1256 p->p_xthread != NULL ?
1257 p->p_xthread->td_tid : -1);
1258 report_alive_proc(td, p, siginfo, status,
1259 options, CLD_TRAPPED);
1263 if ((options & WUNTRACED) != 0 &&
1264 (p->p_flag & P_STOPPED_SIG) != 0) {
1266 report = (p->p_suspcount == p->p_numthreads &&
1267 ((p->p_flag & P_WAITED) == 0));
1270 report_alive_proc(td, p, siginfo, status,
1271 options, CLD_STOPPED);
1275 if ((options & WCONTINUED) != 0 &&
1276 (p->p_flag & P_CONTINUED) != 0) {
1277 report_alive_proc(td, p, siginfo, status, options,
1285 * Look in the orphans list too, to allow the parent to
1286 * collect it's child exit status even if child is being
1289 * Debugger detaches from the parent upon successful
1290 * switch-over from parent to child. At this point due to
1291 * re-parenting the parent loses the child to debugger and a
1292 * wait4(2) call would report that it has no children to wait
1293 * for. By maintaining a list of orphans we allow the parent
1294 * to successfully wait until the child becomes a zombie.
1297 LIST_FOREACH(p, &q->p_orphans, p_orphan) {
1298 ret = proc_to_reap(td, p, idtype, id, NULL, options,
1301 KASSERT(ret != -1, ("reaped an orphan (pid %d)",
1302 (int)td->td_retval[0]));
1310 sx_xunlock(&proctree_lock);
1313 if (options & WNOHANG) {
1314 sx_xunlock(&proctree_lock);
1315 td->td_retval[0] = 0;
1319 if (q->p_flag & P_STATCHILD) {
1320 q->p_flag &= ~P_STATCHILD;
1324 sx_xunlock(&proctree_lock);
1325 error = msleep(q, &q->p_mtx, PWAIT | PCATCH | PDROP, "wait", 0);
1332 proc_add_orphan(struct proc *child, struct proc *parent)
1335 sx_assert(&proctree_lock, SX_XLOCKED);
1336 KASSERT((child->p_flag & P_TRACED) != 0,
1337 ("proc_add_orphan: not traced"));
1339 if (LIST_EMPTY(&parent->p_orphans)) {
1340 child->p_treeflag |= P_TREE_FIRST_ORPHAN;
1341 LIST_INSERT_HEAD(&parent->p_orphans, child, p_orphan);
1343 LIST_INSERT_AFTER(LIST_FIRST(&parent->p_orphans),
1346 child->p_treeflag |= P_TREE_ORPHANED;
1350 * Make process 'parent' the new parent of process 'child'.
1351 * Must be called with an exclusive hold of proctree lock.
1354 proc_reparent(struct proc *child, struct proc *parent, bool set_oppid)
1357 sx_assert(&proctree_lock, SX_XLOCKED);
1358 PROC_LOCK_ASSERT(child, MA_OWNED);
1359 if (child->p_pptr == parent)
1362 PROC_LOCK(child->p_pptr);
1363 sigqueue_take(child->p_ksi);
1364 PROC_UNLOCK(child->p_pptr);
1365 LIST_REMOVE(child, p_sibling);
1366 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
1368 proc_clear_orphan(child);
1369 if ((child->p_flag & P_TRACED) != 0) {
1370 proc_add_orphan(child, child->p_pptr);
1373 child->p_pptr = parent;
1375 child->p_oppid = parent->p_pid;