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 1.54 2006/05/22 06:26:30 swildner 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 sys_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 sysmsg_rundown(lp, 1);
132 caps_exit(lp->lwp_thread);
135 /* are we a task leader? */
136 if(p == p->p_leader) {
137 struct kill_args killArgs;
138 killArgs.signum = SIGKILL;
141 killArgs.pid = q->p_pid;
143 * The interface for kill is better
144 * than the internal signal
151 tsleep((caddr_t)p, 0, "exit1", 0);
157 STOPEVENT(p, S_EXIT, rv);
158 wakeup(&p->p_stype); /* Wakeup anyone in procfs' PIOCWAIT */
161 * Check if any loadable modules need anything done at process exit.
162 * e.g. SYSV IPC stuff
163 * XXX what if one of these generates an error?
165 TAILQ_FOREACH(ep, &exit_list, next)
166 (*ep->function)(p->p_thread);
168 if (p->p_flag & P_PROFIL)
170 MALLOC(p->p_ru, struct rusage *, sizeof(struct rusage),
173 * If parent is waiting for us to exit or exec,
174 * P_PPWAIT is set; we will wakeup the parent below.
176 p->p_flag &= ~(P_TRACED | P_PPWAIT);
177 p->p_flag |= P_WEXIT;
178 SIGEMPTYSET(p->p_siglist);
179 if (timevalisset(&p->p_realtimer.it_value))
180 callout_stop(&p->p_ithandle);
183 * Reset any sigio structures pointing to us as a result of
184 * F_SETOWN with our pid.
186 funsetownlst(&p->p_sigiolst);
189 * Close open files and release open-file table.
195 if(p->p_leader->p_peers) {
197 while(q->p_peers != p)
199 q->p_peers = p->p_peers;
200 wakeup((caddr_t)p->p_leader);
204 * XXX Shutdown SYSV semaphores
208 KKASSERT(p->p_numposixlocks == 0);
210 /* The next two chunks should probably be moved to vmspace_exit. */
214 * Release upcalls associated with this process
217 upc_release(vm, &p->p_lwp);
220 * Release user portion of address space.
221 * This releases references to vnodes,
222 * which could cause I/O if the file has been unlinked.
223 * Need to do this early enough that we can still sleep.
224 * Can't free the entire vmspace as the kernel stack
225 * may be mapped within that space also.
227 * Processes sharing the same vmspace may exit in one order, and
228 * get cleaned up by vmspace_exit() in a different order. The
229 * last exiting process to reach this point releases as much of
230 * the environment as it can, and the last process cleaned up
231 * by vmspace_exit() (which decrements exitingcnt) cleans up the
235 if (--vm->vm_refcnt == 0) {
237 pmap_remove_pages(vmspace_pmap(vm), VM_MIN_ADDRESS,
239 (void) vm_map_remove(&vm->vm_map, VM_MIN_ADDRESS,
243 if (SESS_LEADER(p)) {
244 struct session *sp = p->p_session;
249 * We are the controlling process. Signal the
250 * foreground process group, drain the controlling
251 * terminal, and revoke access to the controlling
254 * NOTE: while waiting for the process group to exit
255 * it is possible that one of the processes in the
256 * group will revoke the tty, so we have to recheck.
258 if (sp->s_ttyp && (sp->s_ttyp->t_session == sp)) {
259 if (sp->s_ttyp->t_pgrp)
260 pgsignal(sp->s_ttyp->t_pgrp, SIGHUP, 1);
261 (void) ttywait(sp->s_ttyp);
263 * The tty could have been revoked
266 if ((vp = sp->s_ttyvp) != NULL) {
267 ttyclosesession(sp, 0);
268 if (vx_lock(vp) == 0) {
269 VOP_REVOKE(vp, REVOKEALL);
272 vrele(vp); /* s_ttyvp ref */
276 * Release the tty. If someone has it open via
277 * /dev/tty then close it (since they no longer can
278 * once we've NULL'd it out).
281 ttyclosesession(sp, 1);
283 * s_ttyp is not zero'd; we use this to indicate
284 * that the session once had a controlling terminal.
285 * (for logging and informational purposes)
290 fixjobc(p, p->p_pgrp, 0);
291 (void)acct_process(p);
297 ktrdestroy(&p->p_tracenode);
301 * Release reference to text vnode
303 if ((vtmp = p->p_textvp) != NULL) {
309 * Once we set SZOMB the process can get reaped. The wait1 code
310 * will also wait for TDF_EXITING to be set and for both TDF_RUNNING
311 * and TDF_PREEMPT_LOCK to be cleared in the thread's flags,
312 * indicating that it has been completely switched out for the last
317 * Remove proc from allproc queue and pidhash chain.
318 * Place onto zombproc. Unlink from parent's child list.
320 * Interlock the SZOMB state with a tsleep against p_lock
321 * (PHOLD/PRELE) so allproc loops don't get confused. Get
322 * our own ref on p_lock to prevent us from getting reaped
325 LIST_REMOVE(p, p_list);
326 LIST_INSERT_HEAD(&zombproc, p, p_list);
327 p->p_flag |= P_ZOMBIE;
329 tsleep(p, 0, "reap1", hz / 10);
330 LIST_REMOVE(p, p_hash);
332 q = LIST_FIRST(&p->p_children);
333 if (q) /* only need this if any child is S_ZOMB */
334 wakeup((caddr_t) initproc);
335 for (; q != 0; q = nq) {
336 nq = LIST_NEXT(q, p_sibling);
337 LIST_REMOVE(q, p_sibling);
338 LIST_INSERT_HEAD(&initproc->p_children, q, p_sibling);
339 q->p_pptr = initproc;
340 q->p_sigparent = SIGCHLD;
342 * Traced processes are killed
343 * since their existence means someone is screwing up.
345 if (q->p_flag & P_TRACED) {
346 q->p_flag &= ~P_TRACED;
352 * Save exit status and final rusage info, adding in child rusage
353 * info and self times.
356 *p->p_ru = p->p_stats->p_ru;
357 calcru(p, &p->p_ru->ru_utime, &p->p_ru->ru_stime, NULL);
358 ruadd(p->p_ru, &p->p_stats->p_cru);
361 * notify interested parties of our demise.
363 KNOTE(&p->p_klist, NOTE_EXIT);
366 * Notify parent that we're gone. If parent has the PS_NOCLDWAIT
367 * flag set, notify process 1 instead (and hope it will handle
370 if (p->p_pptr->p_procsig->ps_flag & PS_NOCLDWAIT) {
371 struct proc *pp = p->p_pptr;
372 proc_reparent(p, initproc);
374 * If this was the last child of our parent, notify
375 * parent, so in case he was wait(2)ing, he will
378 if (LIST_EMPTY(&pp->p_children))
382 if (p->p_sigparent && p->p_pptr != initproc) {
383 psignal(p->p_pptr, p->p_sigparent);
385 psignal(p->p_pptr, SIGCHLD);
388 wakeup((caddr_t)p->p_pptr);
390 * cpu_exit is responsible for clearing curproc, since
391 * it is heavily integrated with the thread/switching sequence.
393 * Other substructures are freed from wait().
395 if (--p->p_limit->p_refcnt == 0) {
396 FREE(p->p_limit, M_SUBPROC);
401 * Release the current user process designation on the process so
402 * the userland scheduler can work in someone else.
404 p->p_usched->release_curproc(lp);
407 * Finally, call machine-dependent code to release the remaining
408 * resources including address space, the kernel stack and pcb.
409 * The address space is released by "vmspace_free(p->p_vmspace)";
410 * This is machine-dependent, as we may have to change stacks
411 * or ensure that the current one isn't reallocated before we
412 * finish. cpu_exit will end with a call to cpu_switch(), finishing
413 * our execution (pun intended).
419 wait4(struct wait_args *uap)
421 struct rusage rusage;
424 error = kern_wait(uap->pid, uap->status ? &status : NULL,
425 uap->options, uap->rusage ? &rusage : NULL, &uap->sysmsg_fds[0]);
427 if (error == 0 && uap->status)
428 error = copyout(&status, uap->status, sizeof(*uap->status));
429 if (error == 0 && uap->rusage)
430 error = copyout(&rusage, uap->rusage, sizeof(*uap->rusage));
437 * wait_args(int pid, int *status, int options, struct rusage *rusage)
440 kern_wait(pid_t pid, int *status, int options, struct rusage *rusage, int *res)
442 struct thread *td = curthread;
443 struct proc *q = td->td_proc;
449 if (options &~ (WUNTRACED|WNOHANG|WLINUXCLONE))
453 * Hack for backwards compatibility with badly written user code.
454 * Or perhaps we have to do this anyway, it is unclear. XXX
456 * The problem is that if a process group is stopped and the parent
457 * is doing a wait*(..., WUNTRACED, ...), it will see the STOP
458 * of the child and then stop itself when it tries to return from the
459 * system call. When the process group is resumed the parent will
460 * then get the STOP status even though the child has now resumed
461 * (a followup wait*() will get the CONT status).
463 * Previously the CONT would overwrite the STOP because the tstop
464 * was handled within tsleep(), and the parent would only see
465 * the CONT when both are stopped and continued together. This litte
466 * two-line hack restores this effect.
468 while (q->p_flag & P_STOPPED)
472 LIST_FOREACH(p, &q->p_children, p_sibling) {
473 if (pid != WAIT_ANY &&
474 p->p_pid != pid && p->p_pgid != -pid)
477 /* This special case handles a kthread spawned by linux_clone
478 * (see linux_misc.c). The linux_wait4 and linux_waitpid
479 * functions need to be able to distinguish between waiting
480 * on a process and waiting on a thread. It is a thread if
481 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option
482 * signifies we want to wait for threads and not processes.
484 if ((p->p_sigparent != SIGCHLD) ^
485 ((options & WLINUXCLONE) != 0)) {
490 if (p->p_flag & P_ZOMBIE) {
492 * Other kernel threads may be in the middle of
493 * accessing the proc. For example, kern/kern_proc.c
494 * could be blocked writing proc data to a sysctl.
495 * At the moment, if this occurs, we are not woken
496 * up and rely on a one-second retry.
500 tsleep(p, 0, "reap3", hz);
502 lwkt_wait_free(p->p_thread);
505 * The process's thread may still be in the middle
506 * of switching away, we can't rip its stack out from
507 * under it until TDF_EXITING is set and both
508 * TDF_RUNNING and TDF_PREEMPT_LOCK are clear.
509 * TDF_PREEMPT_LOCK must be checked because TDF_RUNNING
510 * will be cleared temporarily if a thread gets
513 * YYY no wakeup occurs so we depend on the timeout.
515 if ((p->p_thread->td_flags & (TDF_RUNNING|TDF_PREEMPT_LOCK|TDF_EXITING)) != TDF_EXITING) {
516 tsleep(p->p_thread, 0, "reap2", 1);
520 /* scheduling hook for heuristic */
521 p->p_usched->heuristic_exiting(td->td_lwp, &p->p_lwp);
523 /* Take care of our return values. */
526 *status = p->p_xstat;
530 * If we got the child via a ptrace 'attach',
531 * we need to give it back to the old parent.
533 if (p->p_oppid && (t = pfind(p->p_oppid))) {
541 ruadd(&q->p_stats->p_cru, p->p_ru);
542 FREE(p->p_ru, M_ZOMBIE);
546 * Decrement the count of procs running with this uid.
548 chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0);
551 * Free up credentials.
557 * Remove unused arguments
559 if (p->p_args && --p->p_args->ar_ref == 0)
560 FREE(p->p_args, M_PARGS);
563 * Finally finished with old proc entry.
564 * Unlink it from its process group and free it.
567 LIST_REMOVE(p, p_list); /* off zombproc */
568 LIST_REMOVE(p, p_sibling);
570 if (--p->p_procsig->ps_refcnt == 0) {
571 if (p->p_sigacts != &p->p_addr->u_sigacts)
572 FREE(p->p_sigacts, M_SUBPROC);
573 FREE(p->p_procsig, M_SUBPROC);
582 if ((p->p_flag & P_STOPPED) && (p->p_flag & P_WAITED) == 0 &&
583 (p->p_flag & P_TRACED || options & WUNTRACED)) {
584 p->p_flag |= P_WAITED;
588 *status = W_STOPCODE(p->p_xstat);
589 /* Zero rusage so we get something consistent. */
591 bzero(rusage, sizeof(rusage));
597 if (options & WNOHANG) {
601 error = tsleep((caddr_t)q, PCATCH, "wait", 0);
608 * make process 'parent' the new parent of process 'child'.
611 proc_reparent(struct proc *child, struct proc *parent)
614 if (child->p_pptr == parent)
617 LIST_REMOVE(child, p_sibling);
618 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
619 child->p_pptr = parent;
623 * The next two functions are to handle adding/deleting items on the
627 * Take the arguments given and put them onto the exit callout list,
628 * However first make sure that it's not already there.
629 * returns 0 on success.
633 at_exit(exitlist_fn function)
638 /* Be noisy if the programmer has lost track of things */
639 if (rm_at_exit(function))
640 printf("WARNING: exit callout entry (%p) already present\n",
643 ep = malloc(sizeof(*ep), M_ATEXIT, M_NOWAIT);
646 ep->function = function;
647 TAILQ_INSERT_TAIL(&exit_list, ep, next);
652 * Scan the exit callout list for the given item and remove it.
653 * Returns the number of items removed (0 or 1)
656 rm_at_exit(exitlist_fn function)
660 TAILQ_FOREACH(ep, &exit_list, next) {
661 if (ep->function == function) {
662 TAILQ_REMOVE(&exit_list, ep, next);
673 struct proc *p = curproc;
676 if (p->p_procsig->ps_refcnt == 1 &&
677 p->p_sigacts != &p->p_addr->u_sigacts) {
680 p->p_addr->u_sigacts = *pss;
681 p->p_sigacts = &p->p_addr->u_sigacts;
683 FREE(pss, M_SUBPROC);