thread stage 6: Move thread stack management from the proc structure to
[dragonfly.git] / sys / kern / kern_exit.c
CommitLineData
984263bc
MD
1/*
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.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
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.
25 *
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
36 * SUCH DAMAGE.
37 *
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 $
3020e3be 40 * $DragonFly: src/sys/kern/kern_exit.c,v 1.5 2003/06/19 01:55:06 dillon Exp $
984263bc
MD
41 */
42
43#include "opt_compat.h"
44#include "opt_ktrace.h"
45
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>
51#include <sys/proc.h>
52#include <sys/pioctl.h>
53#include <sys/tty.h>
54#include <sys/wait.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>
61#include <sys/shm.h>
62#include <sys/sem.h>
63#include <sys/aio.h>
64#include <sys/jail.h>
65
66#include <vm/vm.h>
67#include <vm/vm_param.h>
68#include <sys/lock.h>
69#include <vm/pmap.h>
70#include <vm/vm_map.h>
71#include <vm/vm_zone.h>
72#include <vm/vm_extern.h>
73#include <sys/user.h>
74
75/* Required to be non-static for SysVR4 emulator */
76MALLOC_DEFINE(M_ZOMBIE, "zombie", "zombie proc status");
77
78static MALLOC_DEFINE(M_ATEXIT, "atexit", "atexit callback");
79
80static int wait1 __P((struct proc *, struct wait_args *, int));
81
82/*
83 * callout list for things to do at exit time
84 */
85struct exitlist {
86 exitlist_fn function;
87 TAILQ_ENTRY(exitlist) next;
88};
89
90TAILQ_HEAD(exit_list_head, exitlist);
91static struct exit_list_head exit_list = TAILQ_HEAD_INITIALIZER(exit_list);
92
93/*
94 * exit --
95 * Death of process.
96 */
97void
98sys_exit(p, uap)
99 struct proc *p;
100 struct sys_exit_args /* {
101 int rval;
102 } */ *uap;
103{
104
105 exit1(p, W_EXITCODE(uap->rval, 0));
106 /* NOTREACHED */
107}
108
109/*
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.
113 */
114void
115exit1(p, rv)
116 register struct proc *p;
117 int rv;
118{
119 register struct proc *q, *nq;
120 register struct vmspace *vm;
121 struct vnode *vtmp;
122 struct exitlist *ep;
123
124 if (p->p_pid == 1) {
125 printf("init died (signal %d, exit %d)\n",
126 WTERMSIG(rv), WEXITSTATUS(rv));
127 panic("Going nowhere without my init!");
128 }
129
130 aio_proc_rundown(p);
131
132 /* are we a task leader? */
133 if(p == p->p_leader) {
134 struct kill_args killArgs;
135 killArgs.signum = SIGKILL;
136 q = p->p_peers;
137 while(q) {
138 killArgs.pid = q->p_pid;
139 /*
140 * The interface for kill is better
141 * than the internal signal
142 */
143 kill(p, &killArgs);
144 nq = q;
145 q = q->p_peers;
146 }
147 while (p->p_peers)
148 tsleep((caddr_t)p, PWAIT, "exit1", 0);
149 }
150
151#ifdef PGINPROF
152 vmsizmon();
153#endif
154 STOPEVENT(p, S_EXIT, rv);
155 wakeup(&p->p_stype); /* Wakeup anyone in procfs' PIOCWAIT */
156
157 /*
158 * Check if any loadable modules need anything done at process exit.
159 * e.g. SYSV IPC stuff
160 * XXX what if one of these generates an error?
161 */
162 TAILQ_FOREACH(ep, &exit_list, next)
163 (*ep->function)(p);
164
165 if (p->p_flag & P_PROFIL)
166 stopprofclock(p);
167 MALLOC(p->p_ru, struct rusage *, sizeof(struct rusage),
168 M_ZOMBIE, M_WAITOK);
169 /*
170 * If parent is waiting for us to exit or exec,
171 * P_PPWAIT is set; we will wakeup the parent below.
172 */
173 p->p_flag &= ~(P_TRACED | P_PPWAIT);
174 p->p_flag |= P_WEXIT;
175 SIGEMPTYSET(p->p_siglist);
176 if (timevalisset(&p->p_realtimer.it_value))
177 untimeout(realitexpire, (caddr_t)p, p->p_ithandle);
178
179 /*
180 * Reset any sigio structures pointing to us as a result of
181 * F_SETOWN with our pid.
182 */
183 funsetownlst(&p->p_sigiolst);
184
185 /*
186 * Close open files and release open-file table.
187 * This may block!
188 */
189 fdfree(p);
190
191 if(p->p_leader->p_peers) {
192 q = p->p_leader;
193 while(q->p_peers != p)
194 q = q->p_peers;
195 q->p_peers = p->p_peers;
196 wakeup((caddr_t)p->p_leader);
197 }
198
199 /*
200 * XXX Shutdown SYSV semaphores
201 */
202 semexit(p);
203
204 /* The next two chunks should probably be moved to vmspace_exit. */
205 vm = p->p_vmspace;
206 /*
207 * Release user portion of address space.
208 * This releases references to vnodes,
209 * which could cause I/O if the file has been unlinked.
210 * Need to do this early enough that we can still sleep.
211 * Can't free the entire vmspace as the kernel stack
212 * may be mapped within that space also.
213 *
214 * Processes sharing the same vmspace may exit in one order, and
215 * get cleaned up by vmspace_exit() in a different order. The
216 * last exiting process to reach this point releases as much of
217 * the environment as it can, and the last process cleaned up
218 * by vmspace_exit() (which decrements exitingcnt) cleans up the
219 * remainder.
220 */
221 ++vm->vm_exitingcnt;
222 if (--vm->vm_refcnt == 0) {
223 if (vm->vm_shm)
224 shmexit(p);
225 pmap_remove_pages(vmspace_pmap(vm), VM_MIN_ADDRESS,
226 VM_MAXUSER_ADDRESS);
227 (void) vm_map_remove(&vm->vm_map, VM_MIN_ADDRESS,
228 VM_MAXUSER_ADDRESS);
229 }
230
231 if (SESS_LEADER(p)) {
232 register struct session *sp = p->p_session;
233
234 if (sp->s_ttyvp) {
235 /*
236 * Controlling process.
237 * Signal foreground pgrp,
238 * drain controlling terminal
239 * and revoke access to controlling terminal.
240 */
241 if (sp->s_ttyp && (sp->s_ttyp->t_session == sp)) {
242 if (sp->s_ttyp->t_pgrp)
243 pgsignal(sp->s_ttyp->t_pgrp, SIGHUP, 1);
244 (void) ttywait(sp->s_ttyp);
245 /*
246 * The tty could have been revoked
247 * if we blocked.
248 */
249 if (sp->s_ttyvp)
250 VOP_REVOKE(sp->s_ttyvp, REVOKEALL);
251 }
252 if (sp->s_ttyvp)
253 vrele(sp->s_ttyvp);
254 sp->s_ttyvp = NULL;
255 /*
256 * s_ttyp is not zero'd; we use this to indicate
257 * that the session once had a controlling terminal.
258 * (for logging and informational purposes)
259 */
260 }
261 sp->s_leader = NULL;
262 }
263 fixjobc(p, p->p_pgrp, 0);
264 (void)acct_process(p);
265#ifdef KTRACE
266 /*
267 * release trace file
268 */
269 p->p_traceflag = 0; /* don't trace the vrele() */
270 if ((vtmp = p->p_tracep) != NULL) {
271 p->p_tracep = NULL;
272 vrele(vtmp);
273 }
274#endif
275 /*
276 * Release reference to text vnode
277 */
278 if ((vtmp = p->p_textvp) != NULL) {
279 p->p_textvp = NULL;
280 vrele(vtmp);
281 }
282
283 /*
284 * Remove proc from allproc queue and pidhash chain.
285 * Place onto zombproc. Unlink from parent's child list.
286 */
287 LIST_REMOVE(p, p_list);
288 LIST_INSERT_HEAD(&zombproc, p, p_list);
289 p->p_stat = SZOMB;
290
291 LIST_REMOVE(p, p_hash);
292
293 q = LIST_FIRST(&p->p_children);
294 if (q) /* only need this if any child is S_ZOMB */
295 wakeup((caddr_t) initproc);
296 for (; q != 0; q = nq) {
297 nq = LIST_NEXT(q, p_sibling);
298 LIST_REMOVE(q, p_sibling);
299 LIST_INSERT_HEAD(&initproc->p_children, q, p_sibling);
300 q->p_pptr = initproc;
301 q->p_sigparent = SIGCHLD;
302 /*
303 * Traced processes are killed
304 * since their existence means someone is screwing up.
305 */
306 if (q->p_flag & P_TRACED) {
307 q->p_flag &= ~P_TRACED;
308 psignal(q, SIGKILL);
309 }
310 }
311
312 /*
313 * Save exit status and final rusage info, adding in child rusage
314 * info and self times.
315 */
316 p->p_xstat = rv;
317 *p->p_ru = p->p_stats->p_ru;
318 calcru(p, &p->p_ru->ru_utime, &p->p_ru->ru_stime, NULL);
319 ruadd(p->p_ru, &p->p_stats->p_cru);
320
321 /*
322 * Pretend that an mi_switch() to the next process occurs now. We
323 * must set `switchtime' directly since we will call cpu_switch()
324 * directly. Set it now so that the rest of the exit time gets
325 * counted somewhere if possible.
326 */
3020e3be
MD
327 microuptime(&mycpu->gd_switchtime);
328 mycpu->gd_switchticks = ticks;
984263bc
MD
329
330 /*
331 * notify interested parties of our demise.
332 */
333 KNOTE(&p->p_klist, NOTE_EXIT);
334
335 /*
336 * Notify parent that we're gone. If parent has the PS_NOCLDWAIT
337 * flag set, notify process 1 instead (and hope it will handle
338 * this situation).
339 */
340 if (p->p_pptr->p_procsig->ps_flag & PS_NOCLDWAIT) {
341 struct proc *pp = p->p_pptr;
342 proc_reparent(p, initproc);
343 /*
344 * If this was the last child of our parent, notify
345 * parent, so in case he was wait(2)ing, he will
346 * continue.
347 */
348 if (LIST_EMPTY(&pp->p_children))
349 wakeup((caddr_t)pp);
350 }
351
352 if (p->p_sigparent && p->p_pptr != initproc) {
353 psignal(p->p_pptr, p->p_sigparent);
354 } else {
355 psignal(p->p_pptr, SIGCHLD);
356 }
357
358 wakeup((caddr_t)p->p_pptr);
359#if defined(tahoe)
360 /* move this to cpu_exit */
b7c628e4 361 p->p_thread->td_pcb->pcb_saveacc.faddr = (float *)NULL;
984263bc
MD
362#endif
363 /*
364 * Clear curproc after we've done all operations
365 * that could block, and before tearing down the rest
366 * of the process state that might be used from clock, etc.
367 * Also, can't clear curproc while we're still runnable,
368 * as we're not on a run queue (we are current, just not
369 * a proper proc any longer!).
370 *
371 * Other substructures are freed from wait().
372 */
3020e3be 373 mycpu->gd_curthread->td_proc = NULL;
984263bc
MD
374 if (--p->p_limit->p_refcnt == 0) {
375 FREE(p->p_limit, M_SUBPROC);
376 p->p_limit = NULL;
377 }
378
379 /*
380 * Finally, call machine-dependent code to release the remaining
381 * resources including address space, the kernel stack and pcb.
382 * The address space is released by "vmspace_free(p->p_vmspace)";
383 * This is machine-dependent, as we may have to change stacks
384 * or ensure that the current one isn't reallocated before we
385 * finish. cpu_exit will end with a call to cpu_switch(), finishing
386 * our execution (pun intended).
387 */
388 cpu_exit(p);
389}
390
391#ifdef COMPAT_43
392int
393owait(p, uap)
394 struct proc *p;
395 register struct owait_args /* {
396 int dummy;
397 } */ *uap;
398{
399 struct wait_args w;
400
401 w.options = 0;
402 w.rusage = NULL;
403 w.pid = WAIT_ANY;
404 w.status = NULL;
405 return (wait1(p, &w, 1));
406}
407#endif /* COMPAT_43 */
408
409int
410wait4(p, uap)
411 struct proc *p;
412 struct wait_args *uap;
413{
414
415 return (wait1(p, uap, 0));
416}
417
418static int
419wait1(q, uap, compat)
420 register struct proc *q;
421 register struct wait_args /* {
422 int pid;
423 int *status;
424 int options;
425 struct rusage *rusage;
426 } */ *uap;
427 int compat;
428{
429 register int nfound;
430 register struct proc *p, *t;
431 int status, error;
432
433 if (uap->pid == 0)
434 uap->pid = -q->p_pgid;
435 if (uap->options &~ (WUNTRACED|WNOHANG|WLINUXCLONE))
436 return (EINVAL);
437loop:
438 nfound = 0;
439 LIST_FOREACH(p, &q->p_children, p_sibling) {
440 if (uap->pid != WAIT_ANY &&
441 p->p_pid != uap->pid && p->p_pgid != -uap->pid)
442 continue;
443
444 /* This special case handles a kthread spawned by linux_clone
445 * (see linux_misc.c). The linux_wait4 and linux_waitpid functions
446 * need to be able to distinguish between waiting on a process and
447 * waiting on a thread. It is a thread if p_sigparent is not SIGCHLD,
448 * and the WLINUXCLONE option signifies we want to wait for threads
449 * and not processes.
450 */
451 if ((p->p_sigparent != SIGCHLD) ^ ((uap->options & WLINUXCLONE) != 0))
452 continue;
453
454 nfound++;
455 if (p->p_stat == SZOMB) {
456 /* charge childs scheduling cpu usage to parent */
457 if (curproc->p_pid != 1) {
458 curproc->p_estcpu =
459 ESTCPULIM(curproc->p_estcpu + p->p_estcpu);
460 }
461
462 q->p_retval[0] = p->p_pid;
463#ifdef COMPAT_43
464 if (compat)
465 q->p_retval[1] = p->p_xstat;
466 else
467#endif
468 if (uap->status) {
469 status = p->p_xstat; /* convert to int */
470 if ((error = copyout((caddr_t)&status,
471 (caddr_t)uap->status, sizeof(status))))
472 return (error);
473 }
474 if (uap->rusage && (error = copyout((caddr_t)p->p_ru,
475 (caddr_t)uap->rusage, sizeof (struct rusage))))
476 return (error);
477 /*
478 * If we got the child via a ptrace 'attach',
479 * we need to give it back to the old parent.
480 */
481 if (p->p_oppid && (t = pfind(p->p_oppid))) {
482 p->p_oppid = 0;
483 proc_reparent(p, t);
484 psignal(t, SIGCHLD);
485 wakeup((caddr_t)t);
486 return (0);
487 }
488 p->p_xstat = 0;
489 ruadd(&q->p_stats->p_cru, p->p_ru);
490 FREE(p->p_ru, M_ZOMBIE);
491 p->p_ru = NULL;
492
493 /*
494 * Decrement the count of procs running with this uid.
495 */
496 (void)chgproccnt(p->p_cred->p_uidinfo, -1, 0);
497
498 /*
499 * Free up credentials.
500 */
501 if (--p->p_cred->p_refcnt == 0) {
502 crfree(p->p_ucred);
503 uifree(p->p_cred->p_uidinfo);
504 FREE(p->p_cred, M_SUBPROC);
505 p->p_cred = NULL;
506 }
507
508 /*
509 * Destroy empty prisons
510 */
511 if (p->p_prison && !--p->p_prison->pr_ref) {
512 if (p->p_prison->pr_linux != NULL)
513 FREE(p->p_prison->pr_linux, M_PRISON);
514 FREE(p->p_prison, M_PRISON);
515 }
516
517 /*
518 * Remove unused arguments
519 */
520 if (p->p_args && --p->p_args->ar_ref == 0)
521 FREE(p->p_args, M_PARGS);
522
523 /*
524 * Finally finished with old proc entry.
525 * Unlink it from its process group and free it.
526 */
527 leavepgrp(p);
528 LIST_REMOVE(p, p_list); /* off zombproc */
529 LIST_REMOVE(p, p_sibling);
530
531 if (--p->p_procsig->ps_refcnt == 0) {
532 if (p->p_sigacts != &p->p_addr->u_sigacts)
533 FREE(p->p_sigacts, M_SUBPROC);
534 FREE(p->p_procsig, M_SUBPROC);
535 p->p_procsig = NULL;
536 }
537
538 /*
539 * Give machine-dependent layer a chance
540 * to free anything that cpu_exit couldn't
541 * release while still running in process context.
542 */
543 vm_waitproc(p);
544 zfree(proc_zone, p);
545 nprocs--;
546 return (0);
547 }
548 if (p->p_stat == SSTOP && (p->p_flag & P_WAITED) == 0 &&
549 (p->p_flag & P_TRACED || uap->options & WUNTRACED)) {
550 p->p_flag |= P_WAITED;
551 q->p_retval[0] = p->p_pid;
552#ifdef COMPAT_43
553 if (compat) {
554 q->p_retval[1] = W_STOPCODE(p->p_xstat);
555 error = 0;
556 } else
557#endif
558 if (uap->status) {
559 status = W_STOPCODE(p->p_xstat);
560 error = copyout((caddr_t)&status,
561 (caddr_t)uap->status, sizeof(status));
562 } else
563 error = 0;
564 return (error);
565 }
566 }
567 if (nfound == 0)
568 return (ECHILD);
569 if (uap->options & WNOHANG) {
570 q->p_retval[0] = 0;
571 return (0);
572 }
573 if ((error = tsleep((caddr_t)q, PWAIT | PCATCH, "wait", 0)))
574 return (error);
575 goto loop;
576}
577
578/*
579 * make process 'parent' the new parent of process 'child'.
580 */
581void
582proc_reparent(child, parent)
583 register struct proc *child;
584 register struct proc *parent;
585{
586
587 if (child->p_pptr == parent)
588 return;
589
590 LIST_REMOVE(child, p_sibling);
591 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
592 child->p_pptr = parent;
593}
594
595/*
596 * The next two functions are to handle adding/deleting items on the
597 * exit callout list
598 *
599 * at_exit():
600 * Take the arguments given and put them onto the exit callout list,
601 * However first make sure that it's not already there.
602 * returns 0 on success.
603 */
604
605int
606at_exit(function)
607 exitlist_fn function;
608{
609 struct exitlist *ep;
610
611#ifdef INVARIANTS
612 /* Be noisy if the programmer has lost track of things */
613 if (rm_at_exit(function))
614 printf("WARNING: exit callout entry (%p) already present\n",
615 function);
616#endif
617 ep = malloc(sizeof(*ep), M_ATEXIT, M_NOWAIT);
618 if (ep == NULL)
619 return (ENOMEM);
620 ep->function = function;
621 TAILQ_INSERT_TAIL(&exit_list, ep, next);
622 return (0);
623}
624
625/*
626 * Scan the exit callout list for the given item and remove it.
627 * Returns the number of items removed (0 or 1)
628 */
629int
630rm_at_exit(function)
631 exitlist_fn function;
632{
633 struct exitlist *ep;
634
635 TAILQ_FOREACH(ep, &exit_list, next) {
636 if (ep->function == function) {
637 TAILQ_REMOVE(&exit_list, ep, next);
638 free(ep, M_ATEXIT);
639 return(1);
640 }
641 }
642 return (0);
643}
644
645void check_sigacts (void)
646{
647 struct proc *p = curproc;
648 struct sigacts *pss;
649 int s;
650
651 if (p->p_procsig->ps_refcnt == 1 &&
652 p->p_sigacts != &p->p_addr->u_sigacts) {
653 pss = p->p_sigacts;
654 s = splhigh();
655 p->p_addr->u_sigacts = *pss;
656 p->p_sigacts = &p->p_addr->u_sigacts;
657 splx(s);
658 FREE(pss, M_SUBPROC);
659 }
660}
661