FreeBSD/Linux Kernel Cross Reference
sys/kern/kern_exit.c
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 * 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.
21 *
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
32 * SUCH DAMAGE.
33 *
34 * @(#)kern_exit.c 8.7 (Berkeley) 2/12/94
35 */
36
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD$");
39
40 #include "opt_compat.h"
41 #include "opt_kdtrace.h"
42 #include "opt_ktrace.h"
43 #include "opt_mac.h"
44
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/sysproto.h>
48 #include <sys/eventhandler.h>
49 #include <sys/kernel.h>
50 #include <sys/malloc.h>
51 #include <sys/lock.h>
52 #include <sys/mutex.h>
53 #include <sys/proc.h>
54 #include <sys/pioctl.h>
55 #include <sys/tty.h>
56 #include <sys/wait.h>
57 #include <sys/vmmeter.h>
58 #include <sys/vnode.h>
59 #include <sys/resourcevar.h>
60 #include <sys/sbuf.h>
61 #include <sys/signalvar.h>
62 #include <sys/sched.h>
63 #include <sys/sx.h>
64 #include <sys/syscallsubr.h>
65 #include <sys/syslog.h>
66 #include <sys/ptrace.h>
67 #include <sys/acct.h> /* for acct_process() function prototype */
68 #include <sys/filedesc.h>
69 #include <sys/sdt.h>
70 #include <sys/shm.h>
71 #include <sys/sem.h>
72 #ifdef KTRACE
73 #include <sys/ktrace.h>
74 #endif
75
76 #include <security/audit/audit.h>
77 #include <security/mac/mac_framework.h>
78
79 #include <vm/vm.h>
80 #include <vm/vm_extern.h>
81 #include <vm/vm_param.h>
82 #include <vm/pmap.h>
83 #include <vm/vm_map.h>
84 #include <vm/vm_page.h>
85 #include <vm/uma.h>
86
87 #ifdef KDTRACE_HOOKS
88 #include <sys/dtrace_bsd.h>
89 dtrace_execexit_func_t dtrace_fasttrap_exit;
90 #endif
91
92 SDT_PROVIDER_DECLARE(proc);
93 SDT_PROBE_DEFINE(proc, kernel, , exit);
94 SDT_PROBE_ARGTYPE(proc, kernel, , exit, 0, "int");
95
96 /* Required to be non-static for SysVR4 emulator */
97 MALLOC_DEFINE(M_ZOMBIE, "zombie", "zombie proc status");
98
99 /* Hook for NFS teardown procedure. */
100 void (*nlminfo_release_p)(struct proc *p);
101
102 /*
103 * exit -- death of process.
104 */
105 void
106 sys_exit(struct thread *td, struct sys_exit_args *uap)
107 {
108
109 exit1(td, W_EXITCODE(uap->rval, 0));
110 /* NOTREACHED */
111 }
112
113 /*
114 * Exit: deallocate address space and other resources, change proc state to
115 * zombie, and unlink proc from allproc and parent's lists. Save exit status
116 * and rusage for wait(). Check for child processes and orphan them.
117 */
118 void
119 exit1(struct thread *td, int rv)
120 {
121 struct proc *p, *nq, *q;
122 struct tty *tp;
123 struct vnode *ttyvp;
124 struct vnode *vtmp;
125 #ifdef KTRACE
126 struct vnode *tracevp;
127 struct ucred *tracecred;
128 #endif
129 struct plimit *plim;
130 int locked;
131
132 /*
133 * Drop Giant if caller has it. Eventually we should warn about
134 * being called with Giant held.
135 */
136 while (mtx_owned(&Giant))
137 mtx_unlock(&Giant);
138
139 p = td->td_proc;
140 if (p == initproc) {
141 printf("init died (signal %d, exit %d)\n",
142 WTERMSIG(rv), WEXITSTATUS(rv));
143 panic("Going nowhere without my init!");
144 }
145
146 /*
147 * MUST abort all other threads before proceeding past here.
148 */
149 PROC_LOCK(p);
150 if (p->p_flag & P_HADTHREADS) {
151 retry:
152 /*
153 * First check if some other thread got here before us..
154 * if so, act apropriatly, (exit or suspend);
155 */
156 thread_suspend_check(0);
157
158 /*
159 * Kill off the other threads. This requires
160 * some co-operation from other parts of the kernel
161 * so it may not be instantaneous. With this state set
162 * any thread entering the kernel from userspace will
163 * thread_exit() in trap(). Any thread attempting to
164 * sleep will return immediately with EINTR or EWOULDBLOCK
165 * which will hopefully force them to back out to userland
166 * freeing resources as they go. Any thread attempting
167 * to return to userland will thread_exit() from userret().
168 * thread_exit() will unsuspend us when the last of the
169 * other threads exits.
170 * If there is already a thread singler after resumption,
171 * calling thread_single will fail; in that case, we just
172 * re-check all suspension request, the thread should
173 * either be suspended there or exit.
174 */
175 if (thread_single(SINGLE_EXIT))
176 goto retry;
177
178 /*
179 * All other activity in this process is now stopped.
180 * Threading support has been turned off.
181 */
182 }
183 KASSERT(p->p_numthreads == 1,
184 ("exit1: proc %p exiting with %d threads", p, p->p_numthreads));
185 /*
186 * Wakeup anyone in procfs' PIOCWAIT. They should have a hold
187 * on our vmspace, so we should block below until they have
188 * released their reference to us. Note that if they have
189 * requested S_EXIT stops we will block here until they ack
190 * via PIOCCONT.
191 */
192 _STOPEVENT(p, S_EXIT, rv);
193
194 /*
195 * Note that we are exiting and do another wakeup of anyone in
196 * PIOCWAIT in case they aren't listening for S_EXIT stops or
197 * decided to wait again after we told them we are exiting.
198 */
199 p->p_flag |= P_WEXIT;
200 wakeup(&p->p_stype);
201
202 /*
203 * Wait for any processes that have a hold on our vmspace to
204 * release their reference.
205 */
206 while (p->p_lock > 0)
207 msleep(&p->p_lock, &p->p_mtx, PWAIT, "exithold", 0);
208
209 PROC_UNLOCK(p);
210 /* Drain the limit callout while we don't have the proc locked */
211 callout_drain(&p->p_limco);
212
213 #ifdef AUDIT
214 /*
215 * The Sun BSM exit token contains two components: an exit status as
216 * passed to exit(), and a return value to indicate what sort of exit
217 * it was. The exit status is WEXITSTATUS(rv), but it's not clear
218 * what the return value is.
219 */
220 AUDIT_ARG(exit, WEXITSTATUS(rv), 0);
221 AUDIT_SYSCALL_EXIT(0, td);
222 #endif
223
224 /* Are we a task leader? */
225 if (p == p->p_leader) {
226 mtx_lock(&ppeers_lock);
227 q = p->p_peers;
228 while (q != NULL) {
229 PROC_LOCK(q);
230 psignal(q, SIGKILL);
231 PROC_UNLOCK(q);
232 q = q->p_peers;
233 }
234 while (p->p_peers != NULL)
235 msleep(p, &ppeers_lock, PWAIT, "exit1", 0);
236 mtx_unlock(&ppeers_lock);
237 }
238
239 /*
240 * Check if any loadable modules need anything done at process exit.
241 * E.g. SYSV IPC stuff
242 * XXX what if one of these generates an error?
243 */
244 EVENTHANDLER_INVOKE(process_exit, p);
245
246 /*
247 * If parent is waiting for us to exit or exec,
248 * P_PPWAIT is set; we will wakeup the parent below.
249 */
250 PROC_LOCK(p);
251 stopprofclock(p);
252 p->p_flag &= ~(P_TRACED | P_PPWAIT);
253
254 /*
255 * Stop the real interval timer. If the handler is currently
256 * executing, prevent it from rearming itself and let it finish.
257 */
258 if (timevalisset(&p->p_realtimer.it_value) &&
259 callout_stop(&p->p_itcallout) == 0) {
260 timevalclear(&p->p_realtimer.it_interval);
261 msleep(&p->p_itcallout, &p->p_mtx, PWAIT, "ritwait", 0);
262 KASSERT(!timevalisset(&p->p_realtimer.it_value),
263 ("realtime timer is still armed"));
264 }
265 PROC_UNLOCK(p);
266
267 /*
268 * Reset any sigio structures pointing to us as a result of
269 * F_SETOWN with our pid.
270 */
271 funsetownlst(&p->p_sigiolst);
272
273 /*
274 * If this process has an nlminfo data area (for lockd), release it
275 */
276 if (nlminfo_release_p != NULL && p->p_nlminfo != NULL)
277 (*nlminfo_release_p)(p);
278
279 /*
280 * Close open files and release open-file table.
281 * This may block!
282 */
283 fdfree(td);
284
285 /*
286 * If this thread tickled GEOM, we need to wait for the giggling to
287 * stop before we return to userland
288 */
289 if (td->td_pflags & TDP_GEOM)
290 g_waitidle();
291
292 /*
293 * Remove ourself from our leader's peer list and wake our leader.
294 */
295 mtx_lock(&ppeers_lock);
296 if (p->p_leader->p_peers) {
297 q = p->p_leader;
298 while (q->p_peers != p)
299 q = q->p_peers;
300 q->p_peers = p->p_peers;
301 wakeup(p->p_leader);
302 }
303 mtx_unlock(&ppeers_lock);
304
305 vmspace_exit(td);
306
307 mtx_lock(&Giant); /* XXX TTY */
308 sx_xlock(&proctree_lock);
309 if (SESS_LEADER(p)) {
310 struct session *sp;
311
312 sp = p->p_session;
313 if (sp->s_ttyvp) {
314 /*
315 * Controlling process.
316 * Signal foreground pgrp,
317 * drain controlling terminal
318 * and revoke access to controlling terminal.
319 */
320 if (sp->s_ttyp && (sp->s_ttyp->t_session == sp)) {
321 tp = sp->s_ttyp;
322 if (sp->s_ttyp->t_pgrp) {
323 PGRP_LOCK(sp->s_ttyp->t_pgrp);
324 pgsignal(sp->s_ttyp->t_pgrp, SIGHUP, 1);
325 PGRP_UNLOCK(sp->s_ttyp->t_pgrp);
326 }
327 /* XXX tp should be locked. */
328 sx_xunlock(&proctree_lock);
329 (void) ttywait(tp);
330 sx_xlock(&proctree_lock);
331 /*
332 * The tty could have been revoked
333 * if we blocked.
334 */
335 if (sp->s_ttyvp) {
336 ttyvp = sp->s_ttyvp;
337 SESS_LOCK(p->p_session);
338 sp->s_ttyvp = NULL;
339 SESS_UNLOCK(p->p_session);
340 sx_xunlock(&proctree_lock);
341 VOP_LOCK(ttyvp, LK_EXCLUSIVE, td);
342 VOP_REVOKE(ttyvp, REVOKEALL);
343 vput(ttyvp);
344 sx_xlock(&proctree_lock);
345 }
346 }
347 if (sp->s_ttyvp) {
348 ttyvp = sp->s_ttyvp;
349 SESS_LOCK(p->p_session);
350 sp->s_ttyvp = NULL;
351 SESS_UNLOCK(p->p_session);
352 vrele(ttyvp);
353 }
354 /*
355 * s_ttyp is not zero'd; we use this to indicate
356 * that the session once had a controlling terminal.
357 * (for logging and informational purposes)
358 */
359 }
360 SESS_LOCK(p->p_session);
361 sp->s_leader = NULL;
362 SESS_UNLOCK(p->p_session);
363 }
364 fixjobc(p, p->p_pgrp, 0);
365 sx_xunlock(&proctree_lock);
366 (void)acct_process(td);
367 mtx_unlock(&Giant);
368 #ifdef KTRACE
369 /*
370 * Disable tracing, then drain any pending records and release
371 * the trace file.
372 */
373 if (p->p_traceflag != 0) {
374 PROC_LOCK(p);
375 mtx_lock(&ktrace_mtx);
376 p->p_traceflag = 0;
377 mtx_unlock(&ktrace_mtx);
378 PROC_UNLOCK(p);
379 ktrprocexit(td);
380 PROC_LOCK(p);
381 mtx_lock(&ktrace_mtx);
382 tracevp = p->p_tracevp;
383 p->p_tracevp = NULL;
384 tracecred = p->p_tracecred;
385 p->p_tracecred = NULL;
386 mtx_unlock(&ktrace_mtx);
387 PROC_UNLOCK(p);
388 if (tracevp != NULL) {
389 locked = VFS_LOCK_GIANT(tracevp->v_mount);
390 vrele(tracevp);
391 VFS_UNLOCK_GIANT(locked);
392 }
393 if (tracecred != NULL)
394 crfree(tracecred);
395 }
396 #endif
397 /*
398 * Release reference to text vnode
399 */
400 if ((vtmp = p->p_textvp) != NULL) {
401 p->p_textvp = NULL;
402 locked = VFS_LOCK_GIANT(vtmp->v_mount);
403 vrele(vtmp);
404 VFS_UNLOCK_GIANT(locked);
405 }
406
407 /*
408 * Release our limits structure.
409 */
410 PROC_LOCK(p);
411 plim = p->p_limit;
412 p->p_limit = NULL;
413 PROC_UNLOCK(p);
414 lim_free(plim);
415
416 /*
417 * Remove proc from allproc queue and pidhash chain.
418 * Place onto zombproc. Unlink from parent's child list.
419 */
420 sx_xlock(&allproc_lock);
421 LIST_REMOVE(p, p_list);
422 LIST_INSERT_HEAD(&zombproc, p, p_list);
423 LIST_REMOVE(p, p_hash);
424 sx_xunlock(&allproc_lock);
425
426 /*
427 * Call machine-dependent code to release any
428 * machine-dependent resources other than the address space.
429 * The address space is released by "vmspace_exitfree(p)" in
430 * vm_waitproc().
431 */
432 cpu_exit(td);
433
434 WITNESS_WARN(WARN_PANIC, NULL, "process (pid %d) exiting", p->p_pid);
435
436 /*
437 * Reparent all of our children to init.
438 */
439 sx_xlock(&proctree_lock);
440 q = LIST_FIRST(&p->p_children);
441 if (q != NULL) /* only need this if any child is S_ZOMB */
442 wakeup(initproc);
443 for (; q != NULL; q = nq) {
444 nq = LIST_NEXT(q, p_sibling);
445 PROC_LOCK(q);
446 proc_reparent(q, initproc);
447 q->p_sigparent = SIGCHLD;
448 /*
449 * Traced processes are killed
450 * since their existence means someone is screwing up.
451 */
452 if (q->p_flag & P_TRACED) {
453 q->p_flag &= ~(P_TRACED | P_STOPPED_TRACE);
454 psignal(q, SIGKILL);
455 }
456 PROC_UNLOCK(q);
457 }
458
459 /* Save exit status. */
460 PROC_LOCK(p);
461 p->p_xstat = rv;
462 p->p_xthread = td;
463
464 #ifdef KDTRACE_HOOKS
465 /*
466 * Tell the DTrace fasttrap provider about the exit if it
467 * has declared an interest.
468 */
469 if (dtrace_fasttrap_exit)
470 dtrace_fasttrap_exit(p);
471 #endif
472
473 /*
474 * Notify interested parties of our demise.
475 */
476 KNOTE_LOCKED(&p->p_klist, NOTE_EXIT);
477
478 #ifdef KDTRACE_HOOKS
479 int reason = CLD_EXITED;
480 if (WCOREDUMP(rv))
481 reason = CLD_DUMPED;
482 else if (WIFSIGNALED(rv))
483 reason = CLD_KILLED;
484 SDT_PROBE(proc, kernel, , exit, reason, 0, 0, 0, 0);
485 #endif
486 /*
487 * Just delete all entries in the p_klist. At this point we won't
488 * report any more events, and there are nasty race conditions that
489 * can beat us if we don't.
490 */
491 knlist_clear(&p->p_klist, 1);
492
493 /*
494 * Notify parent that we're gone. If parent has the PS_NOCLDWAIT
495 * flag set, or if the handler is set to SIG_IGN, notify process
496 * 1 instead (and hope it will handle this situation).
497 */
498 PROC_LOCK(p->p_pptr);
499 mtx_lock(&p->p_pptr->p_sigacts->ps_mtx);
500 if (p->p_pptr->p_sigacts->ps_flag & (PS_NOCLDWAIT | PS_CLDSIGIGN)) {
501 struct proc *pp;
502
503 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
504 pp = p->p_pptr;
505 PROC_UNLOCK(pp);
506 proc_reparent(p, initproc);
507 p->p_sigparent = SIGCHLD;
508 PROC_LOCK(p->p_pptr);
509 /*
510 * If this was the last child of our parent, notify
511 * parent, so in case he was wait(2)ing, he will
512 * continue.
513 */
514 if (LIST_EMPTY(&pp->p_children))
515 wakeup(pp);
516 } else
517 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
518
519 if (p->p_pptr == initproc)
520 psignal(p->p_pptr, SIGCHLD);
521 else if (p->p_sigparent != 0) {
522 if (p->p_sigparent == SIGCHLD)
523 childproc_exited(p);
524 else /* LINUX thread */
525 psignal(p->p_pptr, p->p_sigparent);
526 }
527 sx_xunlock(&proctree_lock);
528
529 /*
530 * The state PRS_ZOMBIE prevents other proesses from sending
531 * signal to the process, to avoid memory leak, we free memory
532 * for signal queue at the time when the state is set.
533 */
534 sigqueue_flush(&p->p_sigqueue);
535 sigqueue_flush(&td->td_sigqueue);
536
537 /*
538 * We have to wait until after acquiring all locks before
539 * changing p_state. We need to avoid all possible context
540 * switches (including ones from blocking on a mutex) while
541 * marked as a zombie. We also have to set the zombie state
542 * before we release the parent process' proc lock to avoid
543 * a lost wakeup. So, we first call wakeup, then we grab the
544 * sched lock, update the state, and release the parent process'
545 * proc lock.
546 */
547 wakeup(p->p_pptr);
548 PROC_SLOCK(p->p_pptr);
549 sched_exit(p->p_pptr, td);
550 PROC_SUNLOCK(p->p_pptr);
551 PROC_SLOCK(p);
552 p->p_state = PRS_ZOMBIE;
553 PROC_UNLOCK(p->p_pptr);
554
555 /*
556 * Hopefully no one will try to deliver a signal to the process this
557 * late in the game.
558 */
559 knlist_destroy(&p->p_klist);
560
561 /*
562 * Save our children's rusage information in our exit rusage.
563 */
564 ruadd(&p->p_ru, &p->p_rux, &p->p_stats->p_cru, &p->p_crux);
565
566 /*
567 * Make sure the scheduler takes this thread out of its tables etc.
568 * This will also release this thread's reference to the ucred.
569 * Other thread parts to release include pcb bits and such.
570 */
571 thread_exit();
572 }
573
574
575 #ifndef _SYS_SYSPROTO_H_
576 struct abort2_args {
577 char *why;
578 int nargs;
579 void **args;
580 };
581 #endif
582
583 int
584 abort2(struct thread *td, struct abort2_args *uap)
585 {
586 struct proc *p = td->td_proc;
587 struct sbuf *sb;
588 void *uargs[16];
589 int error, i, sig;
590
591 error = 0; /* satisfy compiler */
592
593 /*
594 * Do it right now so we can log either proper call of abort2(), or
595 * note, that invalid argument was passed. 512 is big enough to
596 * handle 16 arguments' descriptions with additional comments.
597 */
598 sb = sbuf_new(NULL, NULL, 512, SBUF_FIXEDLEN);
599 sbuf_clear(sb);
600 sbuf_printf(sb, "%s(pid %d uid %d) aborted: ",
601 p->p_comm, p->p_pid, td->td_ucred->cr_uid);
602 /*
603 * Since we can't return from abort2(), send SIGKILL in cases, where
604 * abort2() was called improperly
605 */
606 sig = SIGKILL;
607 /* Prevent from DoSes from user-space. */
608 if (uap->nargs < 0 || uap->nargs > 16)
609 goto out;
610 if (uap->nargs > 0) {
611 if (uap->args == NULL)
612 goto out;
613 error = copyin(uap->args, uargs, uap->nargs * sizeof(void *));
614 if (error != 0)
615 goto out;
616 }
617 /*
618 * Limit size of 'reason' string to 128. Will fit even when
619 * maximal number of arguments was chosen to be logged.
620 */
621 if (uap->why != NULL) {
622 error = sbuf_copyin(sb, uap->why, 128);
623 if (error < 0)
624 goto out;
625 } else {
626 sbuf_printf(sb, "(null)");
627 }
628 if (uap->nargs > 0) {
629 sbuf_printf(sb, "(");
630 for (i = 0;i < uap->nargs; i++)
631 sbuf_printf(sb, "%s%p", i == 0 ? "" : ", ", uargs[i]);
632 sbuf_printf(sb, ")");
633 }
634 /*
635 * Final stage: arguments were proper, string has been
636 * successfully copied from userspace, and copying pointers
637 * from user-space succeed.
638 */
639 sig = SIGABRT;
640 out:
641 if (sig == SIGKILL) {
642 sbuf_trim(sb);
643 sbuf_printf(sb, " (Reason text inaccessible)");
644 }
645 sbuf_cat(sb, "\n");
646 sbuf_finish(sb);
647 log(LOG_INFO, "%s", sbuf_data(sb));
648 sbuf_delete(sb);
649 exit1(td, W_EXITCODE(0, sig));
650 return (0);
651 }
652
653
654 #ifdef COMPAT_43
655 /*
656 * The dirty work is handled by kern_wait().
657 */
658 int
659 owait(struct thread *td, struct owait_args *uap __unused)
660 {
661 int error, status;
662
663 error = kern_wait(td, WAIT_ANY, &status, 0, NULL);
664 if (error == 0)
665 td->td_retval[1] = status;
666 return (error);
667 }
668 #endif /* COMPAT_43 */
669
670 /*
671 * The dirty work is handled by kern_wait().
672 */
673 int
674 wait4(struct thread *td, struct wait_args *uap)
675 {
676 struct rusage ru, *rup;
677 int error, status;
678
679 if (uap->rusage != NULL)
680 rup = &ru;
681 else
682 rup = NULL;
683 error = kern_wait(td, uap->pid, &status, uap->options, rup);
684 if (uap->status != NULL && error == 0)
685 error = copyout(&status, uap->status, sizeof(status));
686 if (uap->rusage != NULL && error == 0)
687 error = copyout(&ru, uap->rusage, sizeof(struct rusage));
688 return (error);
689 }
690
691 int
692 kern_wait(struct thread *td, pid_t pid, int *status, int options,
693 struct rusage *rusage)
694 {
695 struct proc *p, *q, *t;
696 int error, nfound;
697
698 AUDIT_ARG(pid, pid);
699
700 q = td->td_proc;
701 if (pid == 0) {
702 PROC_LOCK(q);
703 pid = -q->p_pgid;
704 PROC_UNLOCK(q);
705 }
706 if (options &~ (WUNTRACED|WNOHANG|WCONTINUED|WNOWAIT|WLINUXCLONE))
707 return (EINVAL);
708 loop:
709 if (q->p_flag & P_STATCHILD) {
710 PROC_LOCK(q);
711 q->p_flag &= ~P_STATCHILD;
712 PROC_UNLOCK(q);
713 }
714 nfound = 0;
715 sx_xlock(&proctree_lock);
716 LIST_FOREACH(p, &q->p_children, p_sibling) {
717 PROC_LOCK(p);
718 if (pid != WAIT_ANY &&
719 p->p_pid != pid && p->p_pgid != -pid) {
720 PROC_UNLOCK(p);
721 continue;
722 }
723 if (p_canwait(td, p)) {
724 PROC_UNLOCK(p);
725 continue;
726 }
727
728 /*
729 * This special case handles a kthread spawned by linux_clone
730 * (see linux_misc.c). The linux_wait4 and linux_waitpid
731 * functions need to be able to distinguish between waiting
732 * on a process and waiting on a thread. It is a thread if
733 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option
734 * signifies we want to wait for threads and not processes.
735 */
736 if ((p->p_sigparent != SIGCHLD) ^
737 ((options & WLINUXCLONE) != 0)) {
738 PROC_UNLOCK(p);
739 continue;
740 }
741
742 nfound++;
743 PROC_SLOCK(p);
744 if (p->p_state == PRS_ZOMBIE) {
745 if (rusage) {
746 *rusage = p->p_ru;
747 calcru(p, &rusage->ru_utime, &rusage->ru_stime);
748 }
749 PROC_SUNLOCK(p);
750 td->td_retval[0] = p->p_pid;
751 if (status)
752 *status = p->p_xstat; /* convert to int */
753 if (options & WNOWAIT) {
754
755 /*
756 * Only poll, returning the status.
757 * Caller does not wish to release the proc
758 * struct just yet.
759 */
760 PROC_UNLOCK(p);
761 sx_xunlock(&proctree_lock);
762 return (0);
763 }
764
765 PROC_LOCK(q);
766 sigqueue_take(p->p_ksi);
767 PROC_UNLOCK(q);
768 PROC_UNLOCK(p);
769
770 /*
771 * If we got the child via a ptrace 'attach',
772 * we need to give it back to the old parent.
773 */
774 if (p->p_oppid && (t = pfind(p->p_oppid)) != NULL) {
775 PROC_LOCK(p);
776 p->p_oppid = 0;
777 proc_reparent(p, t);
778 PROC_UNLOCK(p);
779 tdsignal(t, NULL, SIGCHLD, p->p_ksi);
780 wakeup(t);
781 PROC_UNLOCK(t);
782 sx_xunlock(&proctree_lock);
783 return (0);
784 }
785
786 /*
787 * Remove other references to this process to ensure
788 * we have an exclusive reference.
789 */
790 sx_xlock(&allproc_lock);
791 LIST_REMOVE(p, p_list); /* off zombproc */
792 sx_xunlock(&allproc_lock);
793 LIST_REMOVE(p, p_sibling);
794 leavepgrp(p);
795 sx_xunlock(&proctree_lock);
796
797 /*
798 * As a side effect of this lock, we know that
799 * all other writes to this proc are visible now, so
800 * no more locking is needed for p.
801 */
802 PROC_LOCK(p);
803 p->p_xstat = 0; /* XXX: why? */
804 PROC_UNLOCK(p);
805 PROC_LOCK(q);
806 ruadd(&q->p_stats->p_cru, &q->p_crux, &p->p_ru,
807 &p->p_rux);
808 PROC_UNLOCK(q);
809
810 /*
811 * Decrement the count of procs running with this uid.
812 */
813 (void)chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0);
814
815 /*
816 * Free credentials, arguments, and sigacts.
817 */
818 crfree(p->p_ucred);
819 p->p_ucred = NULL;
820 pargs_drop(p->p_args);
821 p->p_args = NULL;
822 sigacts_free(p->p_sigacts);
823 p->p_sigacts = NULL;
824
825 /*
826 * Do any thread-system specific cleanups.
827 */
828 thread_wait(p);
829
830 /*
831 * Give vm and machine-dependent layer a chance
832 * to free anything that cpu_exit couldn't
833 * release while still running in process context.
834 */
835 vm_waitproc(p);
836 #ifdef MAC
837 mac_destroy_proc(p);
838 #endif
839 KASSERT(FIRST_THREAD_IN_PROC(p),
840 ("kern_wait: no residual thread!"));
841 uma_zfree(proc_zone, p);
842 sx_xlock(&allproc_lock);
843 nprocs--;
844 sx_xunlock(&allproc_lock);
845 return (0);
846 }
847 if ((p->p_flag & P_STOPPED_SIG) &&
848 (p->p_suspcount == p->p_numthreads) &&
849 (p->p_flag & P_WAITED) == 0 &&
850 (p->p_flag & P_TRACED || options & WUNTRACED)) {
851 PROC_SUNLOCK(p);
852 p->p_flag |= P_WAITED;
853 sx_xunlock(&proctree_lock);
854 td->td_retval[0] = p->p_pid;
855 if (status)
856 *status = W_STOPCODE(p->p_xstat);
857
858 PROC_LOCK(q);
859 sigqueue_take(p->p_ksi);
860 PROC_UNLOCK(q);
861 PROC_UNLOCK(p);
862
863 return (0);
864 }
865 PROC_SUNLOCK(p);
866 if (options & WCONTINUED && (p->p_flag & P_CONTINUED)) {
867 sx_xunlock(&proctree_lock);
868 td->td_retval[0] = p->p_pid;
869 p->p_flag &= ~P_CONTINUED;
870
871 PROC_LOCK(q);
872 sigqueue_take(p->p_ksi);
873 PROC_UNLOCK(q);
874 PROC_UNLOCK(p);
875
876 if (status)
877 *status = SIGCONT;
878 return (0);
879 }
880 PROC_UNLOCK(p);
881 }
882 if (nfound == 0) {
883 sx_xunlock(&proctree_lock);
884 return (ECHILD);
885 }
886 if (options & WNOHANG) {
887 sx_xunlock(&proctree_lock);
888 td->td_retval[0] = 0;
889 return (0);
890 }
891 PROC_LOCK(q);
892 sx_xunlock(&proctree_lock);
893 if (q->p_flag & P_STATCHILD) {
894 q->p_flag &= ~P_STATCHILD;
895 error = 0;
896 } else
897 error = msleep(q, &q->p_mtx, PWAIT | PCATCH, "wait", 0);
898 PROC_UNLOCK(q);
899 if (error)
900 return (error);
901 goto loop;
902 }
903
904 /*
905 * Make process 'parent' the new parent of process 'child'.
906 * Must be called with an exclusive hold of proctree lock.
907 */
908 void
909 proc_reparent(struct proc *child, struct proc *parent)
910 {
911
912 sx_assert(&proctree_lock, SX_XLOCKED);
913 PROC_LOCK_ASSERT(child, MA_OWNED);
914 if (child->p_pptr == parent)
915 return;
916
917 PROC_LOCK(child->p_pptr);
918 sigqueue_take(child->p_ksi);
919 PROC_UNLOCK(child->p_pptr);
920 LIST_REMOVE(child, p_sibling);
921 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
922 child->p_pptr = parent;
923 }
Cache object: 96d760d9a33d51a55c92afadeca0485b
|