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 1.47 2005/10/11 09:59:56 corecode 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/pioctl.h>
55 #include <sys/vnode.h>
56 #include <sys/resourcevar.h>
57 #include <sys/signalvar.h>
58 #include <sys/ptrace.h>
59 #include <sys/acct.h> /* for acct_process() function prototype */
60 #include <sys/filedesc.h>
65 #include <sys/kern_syscall.h>
66 #include <sys/upcall.h>
70 #include <vm/vm_param.h>
73 #include <vm/vm_map.h>
74 #include <vm/vm_zone.h>
75 #include <vm/vm_extern.h>
78 #include <sys/thread2.h>
80 /* Required to be non-static for SysVR4 emulator */
81 MALLOC_DEFINE(M_ZOMBIE, "zombie", "zombie proc status");
83 static MALLOC_DEFINE(M_ATEXIT, "atexit", "atexit callback");
86 * callout list for things to do at exit time
90 TAILQ_ENTRY(exitlist) next;
93 TAILQ_HEAD(exit_list_head, exitlist);
94 static struct exit_list_head exit_list = TAILQ_HEAD_INITIALIZER(exit_list);
100 * SYS_EXIT_ARGS(int rval)
103 sys_exit(struct sys_exit_args *uap)
105 exit1(W_EXITCODE(uap->rval, 0));
110 * Exit: deallocate address space and other resources, change proc state
111 * to zombie, and unlink proc from allproc and parent's lists. Save exit
112 * status and rusage for wait(). Check for child processes and orphan them.
117 struct proc *p = curproc;
125 printf("init died (signal %d, exit %d)\n",
126 WTERMSIG(rv), WEXITSTATUS(rv));
127 panic("Going nowhere without my init!");
130 lp = &p->p_lwp; /* XXX lwp kill other threads */
132 sysmsg_rundown(lp, 1);
133 caps_exit(lp->lwp_thread);
136 /* are we a task leader? */
137 if(p == p->p_leader) {
138 struct kill_args killArgs;
139 killArgs.signum = SIGKILL;
142 killArgs.pid = q->p_pid;
144 * The interface for kill is better
145 * than the internal signal
152 tsleep((caddr_t)p, 0, "exit1", 0);
158 STOPEVENT(p, S_EXIT, rv);
159 wakeup(&p->p_stype); /* Wakeup anyone in procfs' PIOCWAIT */
162 * Check if any loadable modules need anything done at process exit.
163 * e.g. SYSV IPC stuff
164 * XXX what if one of these generates an error?
166 TAILQ_FOREACH(ep, &exit_list, next)
167 (*ep->function)(p->p_thread);
169 if (p->p_flag & P_PROFIL)
171 MALLOC(p->p_ru, struct rusage *, sizeof(struct rusage),
174 * If parent is waiting for us to exit or exec,
175 * P_PPWAIT is set; we will wakeup the parent below.
177 p->p_flag &= ~(P_TRACED | P_PPWAIT);
178 p->p_flag |= P_WEXIT;
179 SIGEMPTYSET(p->p_siglist);
180 if (timevalisset(&p->p_realtimer.it_value))
181 callout_stop(&p->p_ithandle);
184 * Reset any sigio structures pointing to us as a result of
185 * F_SETOWN with our pid.
187 funsetownlst(&p->p_sigiolst);
190 * Close open files and release open-file table.
196 if(p->p_leader->p_peers) {
198 while(q->p_peers != p)
200 q->p_peers = p->p_peers;
201 wakeup((caddr_t)p->p_leader);
205 * XXX Shutdown SYSV semaphores
209 KKASSERT(p->p_numposixlocks == 0);
211 /* The next two chunks should probably be moved to vmspace_exit. */
215 * Release upcalls associated with this process
218 upc_release(vm, &p->p_lwp);
221 * Release user portion of address space.
222 * This releases references to vnodes,
223 * which could cause I/O if the file has been unlinked.
224 * Need to do this early enough that we can still sleep.
225 * Can't free the entire vmspace as the kernel stack
226 * may be mapped within that space also.
228 * Processes sharing the same vmspace may exit in one order, and
229 * get cleaned up by vmspace_exit() in a different order. The
230 * last exiting process to reach this point releases as much of
231 * the environment as it can, and the last process cleaned up
232 * by vmspace_exit() (which decrements exitingcnt) cleans up the
236 if (--vm->vm_refcnt == 0) {
238 pmap_remove_pages(vmspace_pmap(vm), VM_MIN_ADDRESS,
240 (void) vm_map_remove(&vm->vm_map, VM_MIN_ADDRESS,
244 if (SESS_LEADER(p)) {
245 struct session *sp = p->p_session;
250 * We are the controlling process. Signal the
251 * foreground process group, drain the controlling
252 * terminal, and revoke access to the controlling
255 * NOTE: while waiting for the process group to exit
256 * it is possible that one of the processes in the
257 * group will revoke the tty, so we have to recheck.
259 if (sp->s_ttyp && (sp->s_ttyp->t_session == sp)) {
260 if (sp->s_ttyp->t_pgrp)
261 pgsignal(sp->s_ttyp->t_pgrp, SIGHUP, 1);
262 (void) ttywait(sp->s_ttyp);
264 * The tty could have been revoked
267 if ((vp = sp->s_ttyvp) != NULL) {
268 ttyclosesession(sp, 0);
269 if (vx_lock(vp) == 0) {
270 VOP_REVOKE(vp, REVOKEALL);
273 vrele(vp); /* s_ttyvp ref */
277 * Release the tty. If someone has it open via
278 * /dev/tty then close it (since they no longer can
279 * once we've NULL'd it out).
282 ttyclosesession(sp, 1);
284 * s_ttyp is not zero'd; we use this to indicate
285 * that the session once had a controlling terminal.
286 * (for logging and informational purposes)
291 fixjobc(p, p->p_pgrp, 0);
292 (void)acct_process(p);
297 p->p_traceflag = 0; /* don't trace the vrele() */
298 if ((vtmp = p->p_tracep) != NULL) {
304 * Release reference to text vnode
306 if ((vtmp = p->p_textvp) != NULL) {
312 * Once we set SZOMB the process can get reaped. The wait1 code
313 * will also wait for TDF_RUNNING to be cleared in the thread's flags,
314 * indicating that it has been completely switched out.
318 * Remove proc from allproc queue and pidhash chain.
319 * Place onto zombproc. Unlink from parent's child list.
321 LIST_REMOVE(p, p_list);
322 LIST_INSERT_HEAD(&zombproc, p, p_list);
325 LIST_REMOVE(p, p_hash);
327 q = LIST_FIRST(&p->p_children);
328 if (q) /* only need this if any child is S_ZOMB */
329 wakeup((caddr_t) initproc);
330 for (; q != 0; q = nq) {
331 nq = LIST_NEXT(q, p_sibling);
332 LIST_REMOVE(q, p_sibling);
333 LIST_INSERT_HEAD(&initproc->p_children, q, p_sibling);
334 q->p_pptr = initproc;
335 q->p_sigparent = SIGCHLD;
337 * Traced processes are killed
338 * since their existence means someone is screwing up.
340 if (q->p_flag & P_TRACED) {
341 q->p_flag &= ~P_TRACED;
347 * Save exit status and final rusage info, adding in child rusage
348 * info and self times.
351 *p->p_ru = p->p_stats->p_ru;
352 calcru(p, &p->p_ru->ru_utime, &p->p_ru->ru_stime, NULL);
353 ruadd(p->p_ru, &p->p_stats->p_cru);
356 * notify interested parties of our demise.
358 KNOTE(&p->p_klist, NOTE_EXIT);
361 * Notify parent that we're gone. If parent has the PS_NOCLDWAIT
362 * flag set, notify process 1 instead (and hope it will handle
365 if (p->p_pptr->p_procsig->ps_flag & PS_NOCLDWAIT) {
366 struct proc *pp = p->p_pptr;
367 proc_reparent(p, initproc);
369 * If this was the last child of our parent, notify
370 * parent, so in case he was wait(2)ing, he will
373 if (LIST_EMPTY(&pp->p_children))
377 if (p->p_sigparent && p->p_pptr != initproc) {
378 psignal(p->p_pptr, p->p_sigparent);
380 psignal(p->p_pptr, SIGCHLD);
383 wakeup((caddr_t)p->p_pptr);
385 * cpu_exit is responsible for clearing curproc, since
386 * it is heavily integrated with the thread/switching sequence.
388 * Other substructures are freed from wait().
390 if (--p->p_limit->p_refcnt == 0) {
391 FREE(p->p_limit, M_SUBPROC);
396 * Release the current user process designation on the process so
397 * the userland scheduler can work in someone else.
399 p->p_usched->release_curproc(lp);
402 * Finally, call machine-dependent code to release the remaining
403 * resources including address space, the kernel stack and pcb.
404 * The address space is released by "vmspace_free(p->p_vmspace)";
405 * This is machine-dependent, as we may have to change stacks
406 * or ensure that the current one isn't reallocated before we
407 * finish. cpu_exit will end with a call to cpu_switch(), finishing
408 * our execution (pun intended).
414 wait4(struct wait_args *uap)
416 struct rusage rusage;
419 error = kern_wait(uap->pid, uap->status ? &status : NULL,
420 uap->options, uap->rusage ? &rusage : NULL, &uap->sysmsg_fds[0]);
422 if (error == 0 && uap->status)
423 error = copyout(&status, uap->status, sizeof(*uap->status));
424 if (error == 0 && uap->rusage)
425 error = copyout(&rusage, uap->rusage, sizeof(*uap->rusage));
432 * wait_args(int pid, int *status, int options, struct rusage *rusage)
435 kern_wait(pid_t pid, int *status, int options, struct rusage *rusage, int *res)
437 struct thread *td = curthread;
438 struct proc *q = td->td_proc;
444 if (options &~ (WUNTRACED|WNOHANG|WLINUXCLONE))
448 LIST_FOREACH(p, &q->p_children, p_sibling) {
449 if (pid != WAIT_ANY &&
450 p->p_pid != pid && p->p_pgid != -pid)
453 /* This special case handles a kthread spawned by linux_clone
454 * (see linux_misc.c). The linux_wait4 and linux_waitpid
455 * functions need to be able to distinguish between waiting
456 * on a process and waiting on a thread. It is a thread if
457 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option
458 * signifies we want to wait for threads and not processes.
460 if ((p->p_sigparent != SIGCHLD) ^
461 ((options & WLINUXCLONE) != 0)) {
466 if (p->p_stat == SZOMB) {
468 * The process's thread may still be in the middle
469 * of switching away, we can't rip its stack out from
470 * under it until TDF_RUNNING clears!
472 * YYY no wakeup occurs so we depend on the timeout.
474 if ((p->p_thread->td_flags & TDF_RUNNING) != 0) {
475 tsleep(p->p_thread, 0, "reap", 1);
480 * Other kernel threads may be in the middle of
481 * accessing the proc. For example, kern/kern_proc.c
482 * could be blocked writing proc data to a sysctl.
483 * At the moment, if this occurs, we are not woken
484 * up and rely on a one-second retry.
488 tsleep(p, 0, "reap2", hz);
490 lwkt_wait_free(p->p_thread);
492 /* scheduling hook for heuristic */
493 p->p_usched->heuristic_exiting(td->td_lwp, &p->p_lwp);
495 /* Take care of our return values. */
498 *status = p->p_xstat;
502 * If we got the child via a ptrace 'attach',
503 * we need to give it back to the old parent.
505 if (p->p_oppid && (t = pfind(p->p_oppid))) {
513 ruadd(&q->p_stats->p_cru, p->p_ru);
514 FREE(p->p_ru, M_ZOMBIE);
518 * Decrement the count of procs running with this uid.
520 chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0);
523 * Free up credentials.
529 * Remove unused arguments
531 if (p->p_args && --p->p_args->ar_ref == 0)
532 FREE(p->p_args, M_PARGS);
535 * Finally finished with old proc entry.
536 * Unlink it from its process group and free it.
539 LIST_REMOVE(p, p_list); /* off zombproc */
540 LIST_REMOVE(p, p_sibling);
542 if (--p->p_procsig->ps_refcnt == 0) {
543 if (p->p_sigacts != &p->p_addr->u_sigacts)
544 FREE(p->p_sigacts, M_SUBPROC);
545 FREE(p->p_procsig, M_SUBPROC);
554 if (p->p_stat == SSTOP && (p->p_flag & P_WAITED) == 0 &&
555 (p->p_flag & P_TRACED || options & WUNTRACED)) {
556 p->p_flag |= P_WAITED;
560 *status = W_STOPCODE(p->p_xstat);
561 /* Zero rusage so we get something consistent. */
563 bzero(rusage, sizeof(rusage));
569 if (options & WNOHANG) {
573 error = tsleep((caddr_t)q, PCATCH, "wait", 0);
580 * make process 'parent' the new parent of process 'child'.
583 proc_reparent(struct proc *child, struct proc *parent)
586 if (child->p_pptr == parent)
589 LIST_REMOVE(child, p_sibling);
590 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
591 child->p_pptr = parent;
595 * The next two functions are to handle adding/deleting items on the
599 * Take the arguments given and put them onto the exit callout list,
600 * However first make sure that it's not already there.
601 * returns 0 on success.
605 at_exit(exitlist_fn function)
610 /* Be noisy if the programmer has lost track of things */
611 if (rm_at_exit(function))
612 printf("WARNING: exit callout entry (%p) already present\n",
615 ep = malloc(sizeof(*ep), M_ATEXIT, M_NOWAIT);
618 ep->function = function;
619 TAILQ_INSERT_TAIL(&exit_list, ep, next);
624 * Scan the exit callout list for the given item and remove it.
625 * Returns the number of items removed (0 or 1)
628 rm_at_exit(exitlist_fn function)
632 TAILQ_FOREACH(ep, &exit_list, next) {
633 if (ep->function == function) {
634 TAILQ_REMOVE(&exit_list, ep, next);
645 struct proc *p = curproc;
648 if (p->p_procsig->ps_refcnt == 1 &&
649 p->p_sigacts != &p->p_addr->u_sigacts) {
652 p->p_addr->u_sigacts = *pss;
653 p->p_sigacts = &p->p_addr->u_sigacts;
655 FREE(pss, M_SUBPROC);