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"
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/sysproto.h>
47 #include <sys/eventhandler.h>
48 #include <sys/kernel.h>
49 #include <sys/malloc.h>
51 #include <sys/mutex.h>
53 #include <sys/pioctl.h>
56 #include <sys/vmmeter.h>
57 #include <sys/vnode.h>
58 #include <sys/resourcevar.h>
60 #include <sys/signalvar.h>
61 #include <sys/sched.h>
63 #include <sys/syscallsubr.h>
64 #include <sys/syslog.h>
65 #include <sys/ptrace.h>
66 #include <sys/acct.h> /* for acct_process() function prototype */
67 #include <sys/filedesc.h>
71 #include <sys/ktrace.h>
74 #include <security/audit/audit.h>
75 #include <security/mac/mac_framework.h>
78 #include <vm/vm_extern.h>
79 #include <vm/vm_param.h>
81 #include <vm/vm_map.h>
82 #include <vm/vm_page.h>
85 /* Required to be non-static for SysVR4 emulator */
86 MALLOC_DEFINE(M_ZOMBIE, "zombie", "zombie proc status");
88 /* Hook for NFS teardown procedure. */
89 void (*nlminfo_release_p)(struct proc *p);
92 * exit -- death of process.
95 sys_exit(struct thread *td, struct sys_exit_args *uap)
98 exit1(td, W_EXITCODE(uap->rval, 0));
103 * Exit: deallocate address space and other resources, change proc state to
104 * zombie, and unlink proc from allproc and parent's lists. Save exit status
105 * and rusage for wait(). Check for child processes and orphan them.
108 exit1(struct thread *td, int rv)
110 struct proc *p, *nq, *q;
115 struct vnode *tracevp;
116 struct ucred *tracecred;
121 mtx_assert(&Giant, MA_NOTOWNED);
125 printf("init died (signal %d, exit %d)\n",
126 WTERMSIG(rv), WEXITSTATUS(rv));
127 panic("Going nowhere without my init!");
131 * MUST abort all other threads before proceeding past here.
134 while (p->p_flag & P_HADTHREADS) {
136 * First check if some other thread got here before us..
137 * if so, act apropriatly, (exit or suspend);
139 thread_suspend_check(0);
142 * Kill off the other threads. This requires
143 * some co-operation from other parts of the kernel
144 * so it may not be instantaneous. With this state set
145 * any thread entering the kernel from userspace will
146 * thread_exit() in trap(). Any thread attempting to
147 * sleep will return immediately with EINTR or EWOULDBLOCK
148 * which will hopefully force them to back out to userland
149 * freeing resources as they go. Any thread attempting
150 * to return to userland will thread_exit() from userret().
151 * thread_exit() will unsuspend us when the last of the
152 * other threads exits.
153 * If there is already a thread singler after resumption,
154 * calling thread_single will fail; in that case, we just
155 * re-check all suspension request, the thread should
156 * either be suspended there or exit.
158 if (! thread_single(SINGLE_EXIT))
162 * All other activity in this process is now stopped.
163 * Threading support has been turned off.
166 KASSERT(p->p_numthreads == 1,
167 ("exit1: proc %p exiting with %d threads", p, p->p_numthreads));
169 * Wakeup anyone in procfs' PIOCWAIT. They should have a hold
170 * on our vmspace, so we should block below until they have
171 * released their reference to us. Note that if they have
172 * requested S_EXIT stops we will block here until they ack
175 _STOPEVENT(p, S_EXIT, rv);
178 * Note that we are exiting and do another wakeup of anyone in
179 * PIOCWAIT in case they aren't listening for S_EXIT stops or
180 * decided to wait again after we told them we are exiting.
182 p->p_flag |= P_WEXIT;
186 * Wait for any processes that have a hold on our vmspace to
187 * release their reference.
189 while (p->p_lock > 0)
190 msleep(&p->p_lock, &p->p_mtx, PWAIT, "exithold", 0);
193 /* Drain the limit callout while we don't have the proc locked */
194 callout_drain(&p->p_limco);
198 * The Sun BSM exit token contains two components: an exit status as
199 * passed to exit(), and a return value to indicate what sort of exit
200 * it was. The exit status is WEXITSTATUS(rv), but it's not clear
201 * what the return value is.
203 AUDIT_ARG(exit, WEXITSTATUS(rv), 0);
204 AUDIT_SYSCALL_EXIT(0, td);
207 /* Are we a task leader? */
208 if (p == p->p_leader) {
209 mtx_lock(&ppeers_lock);
217 while (p->p_peers != NULL)
218 msleep(p, &ppeers_lock, PWAIT, "exit1", 0);
219 mtx_unlock(&ppeers_lock);
223 * Check if any loadable modules need anything done at process exit.
224 * E.g. SYSV IPC stuff
225 * XXX what if one of these generates an error?
227 EVENTHANDLER_INVOKE(process_exit, p);
230 * If parent is waiting for us to exit or exec,
231 * P_PPWAIT is set; we will wakeup the parent below.
235 p->p_flag &= ~(P_TRACED | P_PPWAIT);
238 * Stop the real interval timer. If the handler is currently
239 * executing, prevent it from rearming itself and let it finish.
241 if (timevalisset(&p->p_realtimer.it_value) &&
242 callout_stop(&p->p_itcallout) == 0) {
243 timevalclear(&p->p_realtimer.it_interval);
244 msleep(&p->p_itcallout, &p->p_mtx, PWAIT, "ritwait", 0);
245 KASSERT(!timevalisset(&p->p_realtimer.it_value),
246 ("realtime timer is still armed"));
251 * Reset any sigio structures pointing to us as a result of
252 * F_SETOWN with our pid.
254 funsetownlst(&p->p_sigiolst);
257 * If this process has an nlminfo data area (for lockd), release it
259 if (nlminfo_release_p != NULL && p->p_nlminfo != NULL)
260 (*nlminfo_release_p)(p);
263 * Close open files and release open-file table.
269 * If this thread tickled GEOM, we need to wait for the giggling to
270 * stop before we return to userland
272 if (td->td_pflags & TDP_GEOM)
276 * Remove ourself from our leader's peer list and wake our leader.
278 mtx_lock(&ppeers_lock);
279 if (p->p_leader->p_peers) {
281 while (q->p_peers != p)
283 q->p_peers = p->p_peers;
286 mtx_unlock(&ppeers_lock);
290 mtx_lock(&Giant); /* XXX TTY */
291 sx_xlock(&proctree_lock);
292 if (SESS_LEADER(p)) {
298 * Controlling process.
299 * Signal foreground pgrp,
300 * drain controlling terminal
301 * and revoke access to controlling terminal.
303 if (sp->s_ttyp && (sp->s_ttyp->t_session == sp)) {
305 if (sp->s_ttyp->t_pgrp) {
306 PGRP_LOCK(sp->s_ttyp->t_pgrp);
307 pgsignal(sp->s_ttyp->t_pgrp, SIGHUP, 1);
308 PGRP_UNLOCK(sp->s_ttyp->t_pgrp);
310 /* XXX tp should be locked. */
311 sx_xunlock(&proctree_lock);
313 sx_xlock(&proctree_lock);
315 * The tty could have been revoked
320 SESS_LOCK(p->p_session);
322 SESS_UNLOCK(p->p_session);
323 sx_xunlock(&proctree_lock);
324 VOP_LOCK(ttyvp, LK_EXCLUSIVE);
325 VOP_REVOKE(ttyvp, REVOKEALL);
327 sx_xlock(&proctree_lock);
332 SESS_LOCK(p->p_session);
334 SESS_UNLOCK(p->p_session);
338 * s_ttyp is not zero'd; we use this to indicate
339 * that the session once had a controlling terminal.
340 * (for logging and informational purposes)
343 SESS_LOCK(p->p_session);
345 SESS_UNLOCK(p->p_session);
347 fixjobc(p, p->p_pgrp, 0);
348 sx_xunlock(&proctree_lock);
349 (void)acct_process(td);
353 * Disable tracing, then drain any pending records and release
356 if (p->p_traceflag != 0) {
358 mtx_lock(&ktrace_mtx);
360 mtx_unlock(&ktrace_mtx);
364 mtx_lock(&ktrace_mtx);
365 tracevp = p->p_tracevp;
367 tracecred = p->p_tracecred;
368 p->p_tracecred = NULL;
369 mtx_unlock(&ktrace_mtx);
371 if (tracevp != NULL) {
372 locked = VFS_LOCK_GIANT(tracevp->v_mount);
374 VFS_UNLOCK_GIANT(locked);
376 if (tracecred != NULL)
381 * Release reference to text vnode
383 if ((vtmp = p->p_textvp) != NULL) {
385 locked = VFS_LOCK_GIANT(vtmp->v_mount);
387 VFS_UNLOCK_GIANT(locked);
391 * Release our limits structure.
400 * Remove proc from allproc queue and pidhash chain.
401 * Place onto zombproc. Unlink from parent's child list.
403 sx_xlock(&allproc_lock);
404 LIST_REMOVE(p, p_list);
405 LIST_INSERT_HEAD(&zombproc, p, p_list);
406 LIST_REMOVE(p, p_hash);
407 sx_xunlock(&allproc_lock);
410 * Call machine-dependent code to release any
411 * machine-dependent resources other than the address space.
412 * The address space is released by "vmspace_exitfree(p)" in
417 WITNESS_WARN(WARN_PANIC, NULL, "process (pid %d) exiting", p->p_pid);
420 * Reparent all of our children to init.
422 sx_xlock(&proctree_lock);
423 q = LIST_FIRST(&p->p_children);
424 if (q != NULL) /* only need this if any child is S_ZOMB */
426 for (; q != NULL; q = nq) {
427 nq = LIST_NEXT(q, p_sibling);
429 proc_reparent(q, initproc);
430 q->p_sigparent = SIGCHLD;
432 * Traced processes are killed
433 * since their existence means someone is screwing up.
435 if (q->p_flag & P_TRACED) {
436 q->p_flag &= ~(P_TRACED | P_STOPPED_TRACE);
442 /* Save exit status. */
447 * Notify interested parties of our demise.
449 KNOTE_LOCKED(&p->p_klist, NOTE_EXIT);
452 * Just delete all entries in the p_klist. At this point we won't
453 * report any more events, and there are nasty race conditions that
454 * can beat us if we don't.
456 knlist_clear(&p->p_klist, 1);
459 * Notify parent that we're gone. If parent has the PS_NOCLDWAIT
460 * flag set, or if the handler is set to SIG_IGN, notify process
461 * 1 instead (and hope it will handle this situation).
463 PROC_LOCK(p->p_pptr);
464 mtx_lock(&p->p_pptr->p_sigacts->ps_mtx);
465 if (p->p_pptr->p_sigacts->ps_flag & (PS_NOCLDWAIT | PS_CLDSIGIGN)) {
468 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
471 proc_reparent(p, initproc);
472 p->p_sigparent = SIGCHLD;
473 PROC_LOCK(p->p_pptr);
475 * If this was the last child of our parent, notify
476 * parent, so in case he was wait(2)ing, he will
479 if (LIST_EMPTY(&pp->p_children))
482 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
484 if (p->p_pptr == initproc)
485 psignal(p->p_pptr, SIGCHLD);
486 else if (p->p_sigparent != 0) {
487 if (p->p_sigparent == SIGCHLD)
489 else /* LINUX thread */
490 psignal(p->p_pptr, p->p_sigparent);
492 sx_xunlock(&proctree_lock);
495 * The state PRS_ZOMBIE prevents other proesses from sending
496 * signal to the process, to avoid memory leak, we free memory
497 * for signal queue at the time when the state is set.
499 sigqueue_flush(&p->p_sigqueue);
500 sigqueue_flush(&td->td_sigqueue);
503 * We have to wait until after acquiring all locks before
504 * changing p_state. We need to avoid all possible context
505 * switches (including ones from blocking on a mutex) while
506 * marked as a zombie. We also have to set the zombie state
507 * before we release the parent process' proc lock to avoid
508 * a lost wakeup. So, we first call wakeup, then we grab the
509 * sched lock, update the state, and release the parent process'
513 sched_exit(p->p_pptr, td);
515 p->p_state = PRS_ZOMBIE;
516 PROC_UNLOCK(p->p_pptr);
519 * Hopefully no one will try to deliver a signal to the process this
522 knlist_destroy(&p->p_klist);
525 * Save our children's rusage information in our exit rusage.
527 ruadd(&p->p_ru, &p->p_rux, &p->p_stats->p_cru, &p->p_crux);
530 * Make sure the scheduler takes this thread out of its tables etc.
531 * This will also release this thread's reference to the ucred.
532 * Other thread parts to release include pcb bits and such.
538 #ifndef _SYS_SYSPROTO_H_
547 abort2(struct thread *td, struct abort2_args *uap)
549 struct proc *p = td->td_proc;
554 error = 0; /* satisfy compiler */
557 * Do it right now so we can log either proper call of abort2(), or
558 * note, that invalid argument was passed. 512 is big enough to
559 * handle 16 arguments' descriptions with additional comments.
561 sb = sbuf_new(NULL, NULL, 512, SBUF_FIXEDLEN);
563 sbuf_printf(sb, "%s(pid %d uid %d) aborted: ",
564 p->p_comm, p->p_pid, td->td_ucred->cr_uid);
566 * Since we can't return from abort2(), send SIGKILL in cases, where
567 * abort2() was called improperly
570 /* Prevent from DoSes from user-space. */
571 if (uap->nargs < 0 || uap->nargs > 16)
573 if (uap->args == NULL)
575 error = copyin(uap->args, uargs, uap->nargs * sizeof(void *));
579 * Limit size of 'reason' string to 128. Will fit even when
580 * maximal number of arguments was chosen to be logged.
582 if (uap->why != NULL) {
583 error = sbuf_copyin(sb, uap->why, 128);
587 sbuf_printf(sb, "(null)");
590 sbuf_printf(sb, "(");
591 for (i = 0;i < uap->nargs; i++)
592 sbuf_printf(sb, "%s%p", i == 0 ? "" : ", ", uargs[i]);
593 sbuf_printf(sb, ")");
596 * Final stage: arguments were proper, string has been
597 * successfully copied from userspace, and copying pointers
598 * from user-space succeed.
602 if (sig == SIGKILL) {
604 sbuf_printf(sb, " (Reason text inaccessible)");
608 log(LOG_INFO, "%s", sbuf_data(sb));
610 exit1(td, W_EXITCODE(0, sig));
617 * The dirty work is handled by kern_wait().
620 owait(struct thread *td, struct owait_args *uap __unused)
624 error = kern_wait(td, WAIT_ANY, &status, 0, NULL);
626 td->td_retval[1] = status;
629 #endif /* COMPAT_43 */
632 * The dirty work is handled by kern_wait().
635 wait4(struct thread *td, struct wait_args *uap)
637 struct rusage ru, *rup;
640 if (uap->rusage != NULL)
644 error = kern_wait(td, uap->pid, &status, uap->options, rup);
645 if (uap->status != NULL && error == 0)
646 error = copyout(&status, uap->status, sizeof(status));
647 if (uap->rusage != NULL && error == 0)
648 error = copyout(&ru, uap->rusage, sizeof(struct rusage));
653 kern_wait(struct thread *td, pid_t pid, int *status, int options,
654 struct rusage *rusage)
656 struct proc *p, *q, *t;
667 if (options &~ (WUNTRACED|WNOHANG|WCONTINUED|WLINUXCLONE))
670 if (q->p_flag & P_STATCHILD) {
672 q->p_flag &= ~P_STATCHILD;
676 sx_xlock(&proctree_lock);
677 LIST_FOREACH(p, &q->p_children, p_sibling) {
679 if (pid != WAIT_ANY &&
680 p->p_pid != pid && p->p_pgid != -pid) {
684 if (p_canwait(td, p)) {
690 * This special case handles a kthread spawned by linux_clone
691 * (see linux_misc.c). The linux_wait4 and linux_waitpid
692 * functions need to be able to distinguish between waiting
693 * on a process and waiting on a thread. It is a thread if
694 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option
695 * signifies we want to wait for threads and not processes.
697 if ((p->p_sigparent != SIGCHLD) ^
698 ((options & WLINUXCLONE) != 0)) {
705 if (p->p_state == PRS_ZOMBIE) {
708 calcru(p, &rusage->ru_utime, &rusage->ru_stime);
711 td->td_retval[0] = p->p_pid;
713 *status = p->p_xstat; /* convert to int */
715 sigqueue_take(p->p_ksi);
719 * If we got the child via a ptrace 'attach',
720 * we need to give it back to the old parent.
723 if (p->p_oppid && (t = pfind(p->p_oppid)) != NULL) {
728 tdsignal(t, NULL, SIGCHLD, p->p_ksi);
731 sx_xunlock(&proctree_lock);
736 * Remove other references to this process to ensure
737 * we have an exclusive reference.
739 sx_xlock(&allproc_lock);
740 LIST_REMOVE(p, p_list); /* off zombproc */
741 sx_xunlock(&allproc_lock);
742 LIST_REMOVE(p, p_sibling);
744 sx_xunlock(&proctree_lock);
747 * As a side effect of this lock, we know that
748 * all other writes to this proc are visible now, so
749 * no more locking is needed for p.
752 p->p_xstat = 0; /* XXX: why? */
755 ruadd(&q->p_stats->p_cru, &q->p_crux, &p->p_ru,
760 * Decrement the count of procs running with this uid.
762 (void)chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0);
765 * Free credentials, arguments, and sigacts.
769 pargs_drop(p->p_args);
771 sigacts_free(p->p_sigacts);
775 * Do any thread-system specific cleanups.
780 * Give vm and machine-dependent layer a chance
781 * to free anything that cpu_exit couldn't
782 * release while still running in process context.
788 KASSERT(FIRST_THREAD_IN_PROC(p),
789 ("kern_wait: no residual thread!"));
790 uma_zfree(proc_zone, p);
791 sx_xlock(&allproc_lock);
793 sx_xunlock(&allproc_lock);
796 if ((p->p_flag & P_STOPPED_SIG) &&
797 (p->p_suspcount == p->p_numthreads) &&
798 (p->p_flag & P_WAITED) == 0 &&
799 (p->p_flag & P_TRACED || options & WUNTRACED)) {
801 p->p_flag |= P_WAITED;
802 sx_xunlock(&proctree_lock);
803 td->td_retval[0] = p->p_pid;
805 *status = W_STOPCODE(p->p_xstat);
808 sigqueue_take(p->p_ksi);
815 if (options & WCONTINUED && (p->p_flag & P_CONTINUED)) {
816 sx_xunlock(&proctree_lock);
817 td->td_retval[0] = p->p_pid;
818 p->p_flag &= ~P_CONTINUED;
821 sigqueue_take(p->p_ksi);
832 sx_xunlock(&proctree_lock);
835 if (options & WNOHANG) {
836 sx_xunlock(&proctree_lock);
837 td->td_retval[0] = 0;
841 sx_xunlock(&proctree_lock);
842 if (q->p_flag & P_STATCHILD) {
843 q->p_flag &= ~P_STATCHILD;
846 error = msleep(q, &q->p_mtx, PWAIT | PCATCH, "wait", 0);
854 * Make process 'parent' the new parent of process 'child'.
855 * Must be called with an exclusive hold of proctree lock.
858 proc_reparent(struct proc *child, struct proc *parent)
861 sx_assert(&proctree_lock, SX_XLOCKED);
862 PROC_LOCK_ASSERT(child, MA_OWNED);
863 if (child->p_pptr == parent)
866 PROC_LOCK(child->p_pptr);
867 sigqueue_take(child->p_ksi);
868 PROC_UNLOCK(child->p_pptr);
869 LIST_REMOVE(child, p_sibling);
870 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
871 child->p_pptr = parent;