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 * 3. All advertising materials mentioning features or use of this software
19 * must display the following acknowledgement:
20 * This product includes software developed by the University of
21 * California, Berkeley and its contributors.
22 * 4. Neither the name of the University nor the names of its contributors
23 * may be used to endorse or promote products derived from this software
24 * without specific prior written permission.
26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
38 * @(#)kern_exit.c 8.7 (Berkeley) 2/12/94
39 * $FreeBSD: src/sys/kern/kern_exit.c,v 1.92.2.11 2003/01/13 22:51:16 dillon Exp $
40 * $DragonFly: src/sys/kern/kern_exit.c,v 1.58 2006/06/07 03:02:10 dillon Exp $
43 #include "opt_compat.h"
44 #include "opt_ktrace.h"
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/sysproto.h>
49 #include <sys/kernel.h>
50 #include <sys/malloc.h>
52 #include <sys/ktrace.h>
53 #include <sys/pioctl.h>
56 #include <sys/vnode.h>
57 #include <sys/resourcevar.h>
58 #include <sys/signalvar.h>
59 #include <sys/ptrace.h>
60 #include <sys/acct.h> /* for acct_process() function prototype */
61 #include <sys/filedesc.h>
66 #include <sys/kern_syscall.h>
67 #include <sys/upcall.h>
71 #include <vm/vm_param.h>
74 #include <vm/vm_map.h>
75 #include <vm/vm_zone.h>
76 #include <vm/vm_extern.h>
79 #include <sys/thread2.h>
81 static MALLOC_DEFINE(M_ATEXIT, "atexit", "atexit callback");
82 static MALLOC_DEFINE(M_ZOMBIE, "zombie", "zombie proc status");
85 * callout list for things to do at exit time
89 TAILQ_ENTRY(exitlist) next;
92 TAILQ_HEAD(exit_list_head, exitlist);
93 static struct exit_list_head exit_list = TAILQ_HEAD_INITIALIZER(exit_list);
99 * SYS_EXIT_ARGS(int rval)
102 sys_exit(struct exit_args *uap)
104 exit1(W_EXITCODE(uap->rval, 0));
109 * Exit: deallocate address space and other resources, change proc state
110 * to zombie, and unlink proc from allproc and parent's lists. Save exit
111 * status and rusage for wait(). Check for child processes and orphan them.
116 struct proc *p = curproc;
124 printf("init died (signal %d, exit %d)\n",
125 WTERMSIG(rv), WEXITSTATUS(rv));
126 panic("Going nowhere without my init!");
129 lp = &p->p_lwp; /* XXX lwp kill other threads */
131 caps_exit(lp->lwp_thread);
134 /* are we a task leader? */
135 if(p == p->p_leader) {
136 struct kill_args killArgs;
137 killArgs.signum = SIGKILL;
140 killArgs.pid = q->p_pid;
142 * The interface for kill is better
143 * than the internal signal
150 tsleep((caddr_t)p, 0, "exit1", 0);
156 STOPEVENT(p, S_EXIT, rv);
157 wakeup(&p->p_stype); /* Wakeup anyone in procfs' PIOCWAIT */
160 * Check if any loadable modules need anything done at process exit.
161 * e.g. SYSV IPC stuff
162 * XXX what if one of these generates an error?
164 TAILQ_FOREACH(ep, &exit_list, next)
165 (*ep->function)(p->p_thread);
167 if (p->p_flag & P_PROFIL)
169 MALLOC(p->p_ru, struct rusage *, sizeof(struct rusage),
172 * If parent is waiting for us to exit or exec,
173 * P_PPWAIT is set; we will wakeup the parent below.
175 p->p_flag &= ~(P_TRACED | P_PPWAIT);
176 p->p_flag |= P_WEXIT;
177 SIGEMPTYSET(p->p_siglist);
178 if (timevalisset(&p->p_realtimer.it_value))
179 callout_stop(&p->p_ithandle);
182 * Reset any sigio structures pointing to us as a result of
183 * F_SETOWN with our pid.
185 funsetownlst(&p->p_sigiolst);
188 * Close open files and release open-file table.
194 if(p->p_leader->p_peers) {
196 while(q->p_peers != p)
198 q->p_peers = p->p_peers;
199 wakeup((caddr_t)p->p_leader);
203 * XXX Shutdown SYSV semaphores
207 KKASSERT(p->p_numposixlocks == 0);
209 /* The next two chunks should probably be moved to vmspace_exit. */
213 * Release upcalls associated with this process
216 upc_release(vm, &p->p_lwp);
219 * Release user portion of address space.
220 * This releases references to vnodes,
221 * which could cause I/O if the file has been unlinked.
222 * Need to do this early enough that we can still sleep.
223 * Can't free the entire vmspace as the kernel stack
224 * may be mapped within that space also.
226 * Processes sharing the same vmspace may exit in one order, and
227 * get cleaned up by vmspace_exit() in a different order. The
228 * last exiting process to reach this point releases as much of
229 * the environment as it can, and the last process cleaned up
230 * by vmspace_exit() (which decrements exitingcnt) cleans up the
234 if (--vm->vm_refcnt == 0) {
236 pmap_remove_pages(vmspace_pmap(vm), VM_MIN_ADDRESS,
238 (void) vm_map_remove(&vm->vm_map, VM_MIN_ADDRESS,
242 if (SESS_LEADER(p)) {
243 struct session *sp = p->p_session;
248 * We are the controlling process. Signal the
249 * foreground process group, drain the controlling
250 * terminal, and revoke access to the controlling
253 * NOTE: while waiting for the process group to exit
254 * it is possible that one of the processes in the
255 * group will revoke the tty, so we have to recheck.
257 if (sp->s_ttyp && (sp->s_ttyp->t_session == sp)) {
258 if (sp->s_ttyp->t_pgrp)
259 pgsignal(sp->s_ttyp->t_pgrp, SIGHUP, 1);
260 (void) ttywait(sp->s_ttyp);
262 * The tty could have been revoked
265 if ((vp = sp->s_ttyvp) != NULL) {
266 ttyclosesession(sp, 0);
267 if (vx_lock(vp) == 0) {
268 VOP_REVOKE(vp, REVOKEALL);
271 vrele(vp); /* s_ttyvp ref */
275 * Release the tty. If someone has it open via
276 * /dev/tty then close it (since they no longer can
277 * once we've NULL'd it out).
280 ttyclosesession(sp, 1);
282 * s_ttyp is not zero'd; we use this to indicate
283 * that the session once had a controlling terminal.
284 * (for logging and informational purposes)
289 fixjobc(p, p->p_pgrp, 0);
290 (void)acct_process(p);
296 ktrdestroy(&p->p_tracenode);
300 * Release reference to text vnode
302 if ((vtmp = p->p_textvp) != NULL) {
308 * Move the process to the zombie list. This will block
309 * until the process p_lock count reaches 0. The process will
310 * not be reaped until TDF_EXITING is set by cpu_thread_exit(),
311 * which is called from cpu_proc_exit().
313 proc_move_allproc_zombie(p);
315 q = LIST_FIRST(&p->p_children);
316 if (q) /* only need this if any child is S_ZOMB */
317 wakeup((caddr_t) initproc);
318 for (; q != 0; q = nq) {
319 nq = LIST_NEXT(q, p_sibling);
320 LIST_REMOVE(q, p_sibling);
321 LIST_INSERT_HEAD(&initproc->p_children, q, p_sibling);
322 q->p_pptr = initproc;
323 q->p_sigparent = SIGCHLD;
325 * Traced processes are killed
326 * since their existence means someone is screwing up.
328 if (q->p_flag & P_TRACED) {
329 q->p_flag &= ~P_TRACED;
335 * Save exit status and final rusage info, adding in child rusage
336 * info and self times.
339 *p->p_ru = p->p_stats->p_ru;
340 calcru(p, &p->p_ru->ru_utime, &p->p_ru->ru_stime, NULL);
341 ruadd(p->p_ru, &p->p_stats->p_cru);
344 * notify interested parties of our demise.
346 KNOTE(&p->p_klist, NOTE_EXIT);
349 * Notify parent that we're gone. If parent has the PS_NOCLDWAIT
350 * flag set, notify process 1 instead (and hope it will handle
353 if (p->p_pptr->p_procsig->ps_flag & PS_NOCLDWAIT) {
354 struct proc *pp = p->p_pptr;
355 proc_reparent(p, initproc);
357 * If this was the last child of our parent, notify
358 * parent, so in case he was wait(2)ing, he will
361 if (LIST_EMPTY(&pp->p_children))
365 if (p->p_sigparent && p->p_pptr != initproc) {
366 psignal(p->p_pptr, p->p_sigparent);
368 psignal(p->p_pptr, SIGCHLD);
371 wakeup((caddr_t)p->p_pptr);
373 * cpu_exit is responsible for clearing curproc, since
374 * it is heavily integrated with the thread/switching sequence.
376 * Other substructures are freed from wait().
378 plimit_free(&p->p_limit);
381 * Release the current user process designation on the process so
382 * the userland scheduler can work in someone else.
384 p->p_usched->release_curproc(lp);
387 * Finally, call machine-dependent code to release the remaining
388 * resources including address space, the kernel stack and pcb.
389 * The address space is released by "vmspace_free(p->p_vmspace)";
390 * This is machine-dependent, as we may have to change stacks
391 * or ensure that the current one isn't reallocated before we
392 * finish. cpu_exit will end with a call to cpu_switch(), finishing
393 * our execution (pun intended).
399 sys_wait4(struct wait_args *uap)
401 struct rusage rusage;
404 error = kern_wait(uap->pid, uap->status ? &status : NULL,
405 uap->options, uap->rusage ? &rusage : NULL, &uap->sysmsg_fds[0]);
407 if (error == 0 && uap->status)
408 error = copyout(&status, uap->status, sizeof(*uap->status));
409 if (error == 0 && uap->rusage)
410 error = copyout(&rusage, uap->rusage, sizeof(*uap->rusage));
417 * wait_args(int pid, int *status, int options, struct rusage *rusage)
420 kern_wait(pid_t pid, int *status, int options, struct rusage *rusage, int *res)
422 struct thread *td = curthread;
423 struct proc *q = td->td_proc;
429 if (options &~ (WUNTRACED|WNOHANG|WLINUXCLONE))
433 * Hack for backwards compatibility with badly written user code.
434 * Or perhaps we have to do this anyway, it is unclear. XXX
436 * The problem is that if a process group is stopped and the parent
437 * is doing a wait*(..., WUNTRACED, ...), it will see the STOP
438 * of the child and then stop itself when it tries to return from the
439 * system call. When the process group is resumed the parent will
440 * then get the STOP status even though the child has now resumed
441 * (a followup wait*() will get the CONT status).
443 * Previously the CONT would overwrite the STOP because the tstop
444 * was handled within tsleep(), and the parent would only see
445 * the CONT when both are stopped and continued together. This litte
446 * two-line hack restores this effect.
448 while (q->p_flag & P_STOPPED)
452 LIST_FOREACH(p, &q->p_children, p_sibling) {
453 if (pid != WAIT_ANY &&
454 p->p_pid != pid && p->p_pgid != -pid)
457 /* This special case handles a kthread spawned by linux_clone
458 * (see linux_misc.c). The linux_wait4 and linux_waitpid
459 * functions need to be able to distinguish between waiting
460 * on a process and waiting on a thread. It is a thread if
461 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option
462 * signifies we want to wait for threads and not processes.
464 if ((p->p_sigparent != SIGCHLD) ^
465 ((options & WLINUXCLONE) != 0)) {
470 if (p->p_flag & P_ZOMBIE) {
472 * Other kernel threads may be in the middle of
473 * accessing the proc. For example, kern/kern_proc.c
474 * could be blocked writing proc data to a sysctl.
475 * At the moment, if this occurs, we are not woken
476 * up and rely on a one-second retry.
480 tsleep(p, 0, "reap3", hz);
482 lwkt_wait_free(p->p_thread);
485 * The process's thread may still be in the middle
486 * of switching away, we can't rip its stack out from
487 * under it until TDF_EXITING is set and both
488 * TDF_RUNNING and TDF_PREEMPT_LOCK are clear.
489 * TDF_PREEMPT_LOCK must be checked because TDF_RUNNING
490 * will be cleared temporarily if a thread gets
493 * YYY no wakeup occurs so we depend on the timeout.
495 if ((p->p_thread->td_flags & (TDF_RUNNING|TDF_PREEMPT_LOCK|TDF_EXITING)) != TDF_EXITING) {
496 tsleep(p->p_thread, 0, "reap2", 1);
500 /* scheduling hook for heuristic */
501 p->p_usched->heuristic_exiting(td->td_lwp, &p->p_lwp);
503 /* Take care of our return values. */
506 *status = p->p_xstat;
510 * If we got the child via a ptrace 'attach',
511 * we need to give it back to the old parent.
513 if (p->p_oppid && (t = pfind(p->p_oppid))) {
521 ruadd(&q->p_stats->p_cru, p->p_ru);
522 FREE(p->p_ru, M_ZOMBIE);
526 * Decrement the count of procs running with this uid.
528 chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0);
531 * Free up credentials.
537 * Remove unused arguments
539 if (p->p_args && --p->p_args->ar_ref == 0)
540 FREE(p->p_args, M_PARGS);
543 * Finally finished with old proc entry.
544 * Unlink it from its process group and free it.
547 proc_remove_zombie(p);
549 if (--p->p_procsig->ps_refcnt == 0) {
550 if (p->p_sigacts != &p->p_addr->u_sigacts)
551 FREE(p->p_sigacts, M_SUBPROC);
552 FREE(p->p_procsig, M_SUBPROC);
561 if ((p->p_flag & P_STOPPED) && (p->p_flag & P_WAITED) == 0 &&
562 (p->p_flag & P_TRACED || options & WUNTRACED)) {
563 p->p_flag |= P_WAITED;
567 *status = W_STOPCODE(p->p_xstat);
568 /* Zero rusage so we get something consistent. */
570 bzero(rusage, sizeof(rusage));
576 if (options & WNOHANG) {
580 error = tsleep((caddr_t)q, PCATCH, "wait", 0);
587 * make process 'parent' the new parent of process 'child'.
590 proc_reparent(struct proc *child, struct proc *parent)
593 if (child->p_pptr == parent)
596 LIST_REMOVE(child, p_sibling);
597 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
598 child->p_pptr = parent;
602 * The next two functions are to handle adding/deleting items on the
606 * Take the arguments given and put them onto the exit callout list,
607 * However first make sure that it's not already there.
608 * returns 0 on success.
612 at_exit(exitlist_fn function)
617 /* Be noisy if the programmer has lost track of things */
618 if (rm_at_exit(function))
619 printf("WARNING: exit callout entry (%p) already present\n",
622 ep = malloc(sizeof(*ep), M_ATEXIT, M_NOWAIT);
625 ep->function = function;
626 TAILQ_INSERT_TAIL(&exit_list, ep, next);
631 * Scan the exit callout list for the given item and remove it.
632 * Returns the number of items removed (0 or 1)
635 rm_at_exit(exitlist_fn function)
639 TAILQ_FOREACH(ep, &exit_list, next) {
640 if (ep->function == function) {
641 TAILQ_REMOVE(&exit_list, ep, next);
652 struct proc *p = curproc;
655 if (p->p_procsig->ps_refcnt == 1 &&
656 p->p_sigacts != &p->p_addr->u_sigacts) {
659 p->p_addr->u_sigacts = *pss;
660 p->p_sigacts = &p->p_addr->u_sigacts;
662 FREE(pss, M_SUBPROC);