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/jail.h>
56 #include <sys/tty.h>
57 #include <sys/wait.h>
58 #include <sys/vmmeter.h>
59 #include <sys/vnode.h>
60 #include <sys/resourcevar.h>
61 #include <sys/sbuf.h>
62 #include <sys/signalvar.h>
63 #include <sys/sched.h>
64 #include <sys/sx.h>
65 #include <sys/syscallsubr.h>
66 #include <sys/syslog.h>
67 #include <sys/ptrace.h>
68 #include <sys/acct.h> /* for acct_process() function prototype */
69 #include <sys/filedesc.h>
70 #include <sys/sdt.h>
71 #include <sys/shm.h>
72 #include <sys/sem.h>
73 #ifdef KTRACE
74 #include <sys/ktrace.h>
75 #endif
76
77 #include <security/audit/audit.h>
78 #include <security/mac/mac_framework.h>
79
80 #include <vm/vm.h>
81 #include <vm/vm_extern.h>
82 #include <vm/vm_param.h>
83 #include <vm/pmap.h>
84 #include <vm/vm_map.h>
85 #include <vm/vm_page.h>
86 #include <vm/uma.h>
87
88 #ifdef KDTRACE_HOOKS
89 #include <sys/dtrace_bsd.h>
90 dtrace_execexit_func_t dtrace_fasttrap_exit;
91 #endif
92
93 SDT_PROVIDER_DECLARE(proc);
94 SDT_PROBE_DEFINE(proc, kernel, , exit);
95 SDT_PROBE_ARGTYPE(proc, kernel, , exit, 0, "int");
96
97 /* Hook for NFS teardown procedure. */
98 void (*nlminfo_release_p)(struct proc *p);
99
100 /*
101 * exit -- death of process.
102 */
103 void
104 sys_exit(struct thread *td, struct sys_exit_args *uap)
105 {
106
107 exit1(td, W_EXITCODE(uap->rval, 0));
108 /* NOTREACHED */
109 }
110
111 /*
112 * Exit: deallocate address space and other resources, change proc state to
113 * zombie, and unlink proc from allproc and parent's lists. Save exit status
114 * and rusage for wait(). Check for child processes and orphan them.
115 */
116 void
117 exit1(struct thread *td, int rv)
118 {
119 struct proc *p, *nq, *q;
120 struct tty *tp;
121 struct vnode *ttyvp;
122 struct vnode *vtmp;
123 struct plimit *plim;
124 int locked;
125
126 /*
127 * Drop Giant if caller has it. Eventually we should warn about
128 * being called with Giant held.
129 */
130 while (mtx_owned(&Giant))
131 mtx_unlock(&Giant);
132
133 p = td->td_proc;
134 if (p == initproc) {
135 printf("init died (signal %d, exit %d)\n",
136 WTERMSIG(rv), WEXITSTATUS(rv));
137 panic("Going nowhere without my init!");
138 }
139
140 /*
141 * MUST abort all other threads before proceeding past here.
142 */
143 PROC_LOCK(p);
144 if (p->p_flag & P_HADTHREADS) {
145 retry:
146 /*
147 * First check if some other thread got here before us..
148 * if so, act apropriatly, (exit or suspend);
149 */
150 thread_suspend_check(0);
151
152 /*
153 * Kill off the other threads. This requires
154 * some co-operation from other parts of the kernel
155 * so it may not be instantaneous. With this state set
156 * any thread entering the kernel from userspace will
157 * thread_exit() in trap(). Any thread attempting to
158 * sleep will return immediately with EINTR or EWOULDBLOCK
159 * which will hopefully force them to back out to userland
160 * freeing resources as they go. Any thread attempting
161 * to return to userland will thread_exit() from userret().
162 * thread_exit() will unsuspend us when the last of the
163 * other threads exits.
164 * If there is already a thread singler after resumption,
165 * calling thread_single will fail; in that case, we just
166 * re-check all suspension request, the thread should
167 * either be suspended there or exit.
168 */
169 if (thread_single(SINGLE_EXIT))
170 goto retry;
171
172 /*
173 * All other activity in this process is now stopped.
174 * Threading support has been turned off.
175 */
176 }
177 KASSERT(p->p_numthreads == 1,
178 ("exit1: proc %p exiting with %d threads", p, p->p_numthreads));
179 /*
180 * Wakeup anyone in procfs' PIOCWAIT. They should have a hold
181 * on our vmspace, so we should block below until they have
182 * released their reference to us. Note that if they have
183 * requested S_EXIT stops we will block here until they ack
184 * via PIOCCONT.
185 */
186 _STOPEVENT(p, S_EXIT, rv);
187
188 /*
189 * Note that we are exiting and do another wakeup of anyone in
190 * PIOCWAIT in case they aren't listening for S_EXIT stops or
191 * decided to wait again after we told them we are exiting.
192 */
193 p->p_flag |= P_WEXIT;
194 wakeup(&p->p_stype);
195
196 /*
197 * Wait for any processes that have a hold on our vmspace to
198 * release their reference.
199 */
200 while (p->p_lock > 0)
201 msleep(&p->p_lock, &p->p_mtx, PWAIT, "exithold", 0);
202
203 PROC_UNLOCK(p);
204 /* Drain the limit callout while we don't have the proc locked */
205 callout_drain(&p->p_limco);
206
207 #ifdef AUDIT
208 /*
209 * The Sun BSM exit token contains two components: an exit status as
210 * passed to exit(), and a return value to indicate what sort of exit
211 * it was. The exit status is WEXITSTATUS(rv), but it's not clear
212 * what the return value is.
213 */
214 AUDIT_ARG(exit, WEXITSTATUS(rv), 0);
215 AUDIT_SYSCALL_EXIT(0, td);
216 #endif
217
218 /* Are we a task leader? */
219 if (p == p->p_leader) {
220 mtx_lock(&ppeers_lock);
221 q = p->p_peers;
222 while (q != NULL) {
223 PROC_LOCK(q);
224 psignal(q, SIGKILL);
225 PROC_UNLOCK(q);
226 q = q->p_peers;
227 }
228 while (p->p_peers != NULL)
229 msleep(p, &ppeers_lock, PWAIT, "exit1", 0);
230 mtx_unlock(&ppeers_lock);
231 }
232
233 /*
234 * Check if any loadable modules need anything done at process exit.
235 * E.g. SYSV IPC stuff
236 * XXX what if one of these generates an error?
237 */
238 EVENTHANDLER_INVOKE(process_exit, p);
239
240 /*
241 * If parent is waiting for us to exit or exec,
242 * P_PPWAIT is set; we will wakeup the parent below.
243 */
244 PROC_LOCK(p);
245 stopprofclock(p);
246 p->p_flag &= ~(P_TRACED | P_PPWAIT);
247
248 /*
249 * Stop the real interval timer. If the handler is currently
250 * executing, prevent it from rearming itself and let it finish.
251 */
252 if (timevalisset(&p->p_realtimer.it_value) &&
253 callout_stop(&p->p_itcallout) == 0) {
254 timevalclear(&p->p_realtimer.it_interval);
255 msleep(&p->p_itcallout, &p->p_mtx, PWAIT, "ritwait", 0);
256 KASSERT(!timevalisset(&p->p_realtimer.it_value),
257 ("realtime timer is still armed"));
258 }
259 PROC_UNLOCK(p);
260
261 /*
262 * Reset any sigio structures pointing to us as a result of
263 * F_SETOWN with our pid.
264 */
265 funsetownlst(&p->p_sigiolst);
266
267 /*
268 * If this process has an nlminfo data area (for lockd), release it
269 */
270 if (nlminfo_release_p != NULL && p->p_nlminfo != NULL)
271 (*nlminfo_release_p)(p);
272
273 /*
274 * Close open files and release open-file table.
275 * This may block!
276 */
277 fdfree(td);
278
279 /*
280 * If this thread tickled GEOM, we need to wait for the giggling to
281 * stop before we return to userland
282 */
283 if (td->td_pflags & TDP_GEOM)
284 g_waitidle();
285
286 /*
287 * Remove ourself from our leader's peer list and wake our leader.
288 */
289 mtx_lock(&ppeers_lock);
290 if (p->p_leader->p_peers) {
291 q = p->p_leader;
292 while (q->p_peers != p)
293 q = q->p_peers;
294 q->p_peers = p->p_peers;
295 wakeup(p->p_leader);
296 }
297 mtx_unlock(&ppeers_lock);
298
299 vmspace_exit(td);
300
301 mtx_lock(&Giant); /* XXX TTY */
302 sx_xlock(&proctree_lock);
303 if (SESS_LEADER(p)) {
304 struct session *sp;
305
306 sp = p->p_session;
307 if (sp->s_ttyvp) {
308 /*
309 * Controlling process.
310 * Signal foreground pgrp,
311 * drain controlling terminal
312 * and revoke access to controlling terminal.
313 */
314 if (sp->s_ttyp && (sp->s_ttyp->t_session == sp)) {
315 tp = sp->s_ttyp;
316 if (sp->s_ttyp->t_pgrp) {
317 PGRP_LOCK(sp->s_ttyp->t_pgrp);
318 pgsignal(sp->s_ttyp->t_pgrp, SIGHUP, 1);
319 PGRP_UNLOCK(sp->s_ttyp->t_pgrp);
320 }
321 /* XXX tp should be locked. */
322 sx_xunlock(&proctree_lock);
323 (void) ttywait(tp);
324 sx_xlock(&proctree_lock);
325 /*
326 * The tty could have been revoked
327 * if we blocked.
328 */
329 if (sp->s_ttyvp) {
330 ttyvp = sp->s_ttyvp;
331 SESS_LOCK(p->p_session);
332 sp->s_ttyvp = NULL;
333 SESS_UNLOCK(p->p_session);
334 sx_xunlock(&proctree_lock);
335 VOP_LOCK(ttyvp, LK_EXCLUSIVE, td);
336 VOP_REVOKE(ttyvp, REVOKEALL);
337 vput(ttyvp);
338 sx_xlock(&proctree_lock);
339 }
340 }
341 if (sp->s_ttyvp) {
342 ttyvp = sp->s_ttyvp;
343 SESS_LOCK(p->p_session);
344 sp->s_ttyvp = NULL;
345 SESS_UNLOCK(p->p_session);
346 vrele(ttyvp);
347 }
348 /*
349 * s_ttyp is not zero'd; we use this to indicate
350 * that the session once had a controlling terminal.
351 * (for logging and informational purposes)
352 */
353 }
354 SESS_LOCK(p->p_session);
355 sp->s_leader = NULL;
356 SESS_UNLOCK(p->p_session);
357 }
358 fixjobc(p, p->p_pgrp, 0);
359 sx_xunlock(&proctree_lock);
360 (void)acct_process(td);
361 mtx_unlock(&Giant);
362 #ifdef KTRACE
363 ktrprocexit(td);
364 #endif
365 /*
366 * Release reference to text vnode
367 */
368 if ((vtmp = p->p_textvp) != NULL) {
369 p->p_textvp = NULL;
370 locked = VFS_LOCK_GIANT(vtmp->v_mount);
371 vrele(vtmp);
372 VFS_UNLOCK_GIANT(locked);
373 }
374
375 /*
376 * Release our limits structure.
377 */
378 PROC_LOCK(p);
379 plim = p->p_limit;
380 p->p_limit = NULL;
381 PROC_UNLOCK(p);
382 lim_free(plim);
383
384 /*
385 * Remove proc from allproc queue and pidhash chain.
386 * Place onto zombproc. Unlink from parent's child list.
387 */
388 sx_xlock(&allproc_lock);
389 LIST_REMOVE(p, p_list);
390 LIST_INSERT_HEAD(&zombproc, p, p_list);
391 LIST_REMOVE(p, p_hash);
392 sx_xunlock(&allproc_lock);
393
394 /*
395 * Call machine-dependent code to release any
396 * machine-dependent resources other than the address space.
397 * The address space is released by "vmspace_exitfree(p)" in
398 * vm_waitproc().
399 */
400 cpu_exit(td);
401
402 WITNESS_WARN(WARN_PANIC, NULL, "process (pid %d) exiting", p->p_pid);
403
404 /*
405 * Reparent all of our children to init.
406 */
407 sx_xlock(&proctree_lock);
408 q = LIST_FIRST(&p->p_children);
409 if (q != NULL) /* only need this if any child is S_ZOMB */
410 wakeup(initproc);
411 for (; q != NULL; q = nq) {
412 nq = LIST_NEXT(q, p_sibling);
413 PROC_LOCK(q);
414 proc_reparent(q, initproc);
415 q->p_sigparent = SIGCHLD;
416 /*
417 * Traced processes are killed
418 * since their existence means someone is screwing up.
419 */
420 if (q->p_flag & P_TRACED) {
421 q->p_flag &= ~(P_TRACED | P_STOPPED_TRACE);
422 psignal(q, SIGKILL);
423 }
424 PROC_UNLOCK(q);
425 }
426
427 /* Save exit status. */
428 PROC_LOCK(p);
429 p->p_xstat = rv;
430 p->p_xthread = td;
431
432 /* In case we are jailed tell the prison that we are gone. */
433 if (jailed(p->p_ucred))
434 prison_proc_free(p->p_ucred->cr_prison);
435
436 #ifdef KDTRACE_HOOKS
437 /*
438 * Tell the DTrace fasttrap provider about the exit if it
439 * has declared an interest.
440 */
441 if (dtrace_fasttrap_exit)
442 dtrace_fasttrap_exit(p);
443 #endif
444
445 /*
446 * Notify interested parties of our demise.
447 */
448 KNOTE_LOCKED(&p->p_klist, NOTE_EXIT);
449
450 #ifdef KDTRACE_HOOKS
451 int reason = CLD_EXITED;
452 if (WCOREDUMP(rv))
453 reason = CLD_DUMPED;
454 else if (WIFSIGNALED(rv))
455 reason = CLD_KILLED;
456 SDT_PROBE(proc, kernel, , exit, reason, 0, 0, 0, 0);
457 #endif
458 /*
459 * Just delete all entries in the p_klist. At this point we won't
460 * report any more events, and there are nasty race conditions that
461 * can beat us if we don't.
462 */
463 knlist_clear(&p->p_klist, 1);
464
465 /*
466 * Notify parent that we're gone. If parent has the PS_NOCLDWAIT
467 * flag set, or if the handler is set to SIG_IGN, notify process
468 * 1 instead (and hope it will handle this situation).
469 */
470 PROC_LOCK(p->p_pptr);
471 mtx_lock(&p->p_pptr->p_sigacts->ps_mtx);
472 if (p->p_pptr->p_sigacts->ps_flag & (PS_NOCLDWAIT | PS_CLDSIGIGN)) {
473 struct proc *pp;
474
475 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
476 pp = p->p_pptr;
477 PROC_UNLOCK(pp);
478 proc_reparent(p, initproc);
479 p->p_sigparent = SIGCHLD;
480 PROC_LOCK(p->p_pptr);
481
482 /*
483 * Notify parent, so in case he was wait(2)ing or
484 * executing waitpid(2) with our pid, he will
485 * continue.
486 */
487 wakeup(pp);
488 } else
489 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
490
491 if (p->p_pptr == initproc)
492 psignal(p->p_pptr, SIGCHLD);
493 else if (p->p_sigparent != 0) {
494 if (p->p_sigparent == SIGCHLD)
495 childproc_exited(p);
496 else /* LINUX thread */
497 psignal(p->p_pptr, p->p_sigparent);
498 }
499 sx_xunlock(&proctree_lock);
500
501 /*
502 * The state PRS_ZOMBIE prevents other proesses from sending
503 * signal to the process, to avoid memory leak, we free memory
504 * for signal queue at the time when the state is set.
505 */
506 sigqueue_flush(&p->p_sigqueue);
507 sigqueue_flush(&td->td_sigqueue);
508
509 /*
510 * We have to wait until after acquiring all locks before
511 * changing p_state. We need to avoid all possible context
512 * switches (including ones from blocking on a mutex) while
513 * marked as a zombie. We also have to set the zombie state
514 * before we release the parent process' proc lock to avoid
515 * a lost wakeup. So, we first call wakeup, then we grab the
516 * sched lock, update the state, and release the parent process'
517 * proc lock.
518 */
519 wakeup(p->p_pptr);
520 cv_broadcast(&p->p_pwait);
521 PROC_SLOCK(p->p_pptr);
522 sched_exit(p->p_pptr, td);
523 PROC_SUNLOCK(p->p_pptr);
524 PROC_SLOCK(p);
525 p->p_state = PRS_ZOMBIE;
526 PROC_UNLOCK(p->p_pptr);
527
528 /*
529 * Hopefully no one will try to deliver a signal to the process this
530 * late in the game.
531 */
532 knlist_destroy(&p->p_klist);
533
534 /*
535 * Save our children's rusage information in our exit rusage.
536 */
537 ruadd(&p->p_ru, &p->p_rux, &p->p_stats->p_cru, &p->p_crux);
538
539 /*
540 * Make sure the scheduler takes this thread out of its tables etc.
541 * This will also release this thread's reference to the ucred.
542 * Other thread parts to release include pcb bits and such.
543 */
544 thread_exit();
545 }
546
547
548 #ifndef _SYS_SYSPROTO_H_
549 struct abort2_args {
550 char *why;
551 int nargs;
552 void **args;
553 };
554 #endif
555
556 int
557 abort2(struct thread *td, struct abort2_args *uap)
558 {
559 struct proc *p = td->td_proc;
560 struct sbuf *sb;
561 void *uargs[16];
562 int error, i, sig;
563
564 error = 0; /* satisfy compiler */
565
566 /*
567 * Do it right now so we can log either proper call of abort2(), or
568 * note, that invalid argument was passed. 512 is big enough to
569 * handle 16 arguments' descriptions with additional comments.
570 */
571 sb = sbuf_new(NULL, NULL, 512, SBUF_FIXEDLEN);
572 sbuf_clear(sb);
573 sbuf_printf(sb, "%s(pid %d uid %d) aborted: ",
574 p->p_comm, p->p_pid, td->td_ucred->cr_uid);
575 /*
576 * Since we can't return from abort2(), send SIGKILL in cases, where
577 * abort2() was called improperly
578 */
579 sig = SIGKILL;
580 /* Prevent from DoSes from user-space. */
581 if (uap->nargs < 0 || uap->nargs > 16)
582 goto out;
583 if (uap->nargs > 0) {
584 if (uap->args == NULL)
585 goto out;
586 error = copyin(uap->args, uargs, uap->nargs * sizeof(void *));
587 if (error != 0)
588 goto out;
589 }
590 /*
591 * Limit size of 'reason' string to 128. Will fit even when
592 * maximal number of arguments was chosen to be logged.
593 */
594 if (uap->why != NULL) {
595 error = sbuf_copyin(sb, uap->why, 128);
596 if (error < 0)
597 goto out;
598 } else {
599 sbuf_printf(sb, "(null)");
600 }
601 if (uap->nargs > 0) {
602 sbuf_printf(sb, "(");
603 for (i = 0;i < uap->nargs; i++)
604 sbuf_printf(sb, "%s%p", i == 0 ? "" : ", ", uargs[i]);
605 sbuf_printf(sb, ")");
606 }
607 /*
608 * Final stage: arguments were proper, string has been
609 * successfully copied from userspace, and copying pointers
610 * from user-space succeed.
611 */
612 sig = SIGABRT;
613 out:
614 if (sig == SIGKILL) {
615 sbuf_trim(sb);
616 sbuf_printf(sb, " (Reason text inaccessible)");
617 }
618 sbuf_cat(sb, "\n");
619 sbuf_finish(sb);
620 log(LOG_INFO, "%s", sbuf_data(sb));
621 sbuf_delete(sb);
622 exit1(td, W_EXITCODE(0, sig));
623 return (0);
624 }
625
626
627 #ifdef COMPAT_43
628 /*
629 * The dirty work is handled by kern_wait().
630 */
631 int
632 owait(struct thread *td, struct owait_args *uap __unused)
633 {
634 int error, status;
635
636 error = kern_wait(td, WAIT_ANY, &status, 0, NULL);
637 if (error == 0)
638 td->td_retval[1] = status;
639 return (error);
640 }
641 #endif /* COMPAT_43 */
642
643 /*
644 * The dirty work is handled by kern_wait().
645 */
646 int
647 wait4(struct thread *td, struct wait_args *uap)
648 {
649 struct rusage ru, *rup;
650 int error, status;
651
652 if (uap->rusage != NULL)
653 rup = &ru;
654 else
655 rup = NULL;
656 error = kern_wait(td, uap->pid, &status, uap->options, rup);
657 if (uap->status != NULL && error == 0)
658 error = copyout(&status, uap->status, sizeof(status));
659 if (uap->rusage != NULL && error == 0)
660 error = copyout(&ru, uap->rusage, sizeof(struct rusage));
661 return (error);
662 }
663
664 int
665 kern_wait(struct thread *td, pid_t pid, int *status, int options,
666 struct rusage *rusage)
667 {
668 struct proc *p, *q, *t;
669 int error, nfound;
670
671 AUDIT_ARG(pid, pid);
672
673 q = td->td_proc;
674 if (pid == 0) {
675 PROC_LOCK(q);
676 pid = -q->p_pgid;
677 PROC_UNLOCK(q);
678 }
679 if (options &~ (WUNTRACED|WNOHANG|WCONTINUED|WNOWAIT|WLINUXCLONE))
680 return (EINVAL);
681 loop:
682 if (q->p_flag & P_STATCHILD) {
683 PROC_LOCK(q);
684 q->p_flag &= ~P_STATCHILD;
685 PROC_UNLOCK(q);
686 }
687 nfound = 0;
688 sx_xlock(&proctree_lock);
689 LIST_FOREACH(p, &q->p_children, p_sibling) {
690 PROC_LOCK(p);
691 if (pid != WAIT_ANY &&
692 p->p_pid != pid && p->p_pgid != -pid) {
693 PROC_UNLOCK(p);
694 continue;
695 }
696 if (p_canwait(td, p)) {
697 PROC_UNLOCK(p);
698 continue;
699 }
700
701 /*
702 * This special case handles a kthread spawned by linux_clone
703 * (see linux_misc.c). The linux_wait4 and linux_waitpid
704 * functions need to be able to distinguish between waiting
705 * on a process and waiting on a thread. It is a thread if
706 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option
707 * signifies we want to wait for threads and not processes.
708 */
709 if ((p->p_sigparent != SIGCHLD) ^
710 ((options & WLINUXCLONE) != 0)) {
711 PROC_UNLOCK(p);
712 continue;
713 }
714
715 nfound++;
716 PROC_SLOCK(p);
717 if (p->p_state == PRS_ZOMBIE) {
718 if (rusage) {
719 *rusage = p->p_ru;
720 calcru(p, &rusage->ru_utime, &rusage->ru_stime);
721 }
722 PROC_SUNLOCK(p);
723 td->td_retval[0] = p->p_pid;
724 if (status)
725 *status = p->p_xstat; /* convert to int */
726 if (options & WNOWAIT) {
727
728 /*
729 * Only poll, returning the status.
730 * Caller does not wish to release the proc
731 * struct just yet.
732 */
733 PROC_UNLOCK(p);
734 sx_xunlock(&proctree_lock);
735 return (0);
736 }
737
738 PROC_LOCK(q);
739 sigqueue_take(p->p_ksi);
740 PROC_UNLOCK(q);
741 PROC_UNLOCK(p);
742
743 /*
744 * If we got the child via a ptrace 'attach',
745 * we need to give it back to the old parent.
746 */
747 if (p->p_oppid && (t = pfind(p->p_oppid)) != NULL) {
748 PROC_LOCK(p);
749 p->p_oppid = 0;
750 proc_reparent(p, t);
751 PROC_UNLOCK(p);
752 tdsignal(t, NULL, SIGCHLD, p->p_ksi);
753 wakeup(t);
754 cv_broadcast(&p->p_pwait);
755 PROC_UNLOCK(t);
756 sx_xunlock(&proctree_lock);
757 return (0);
758 }
759
760 /*
761 * Remove other references to this process to ensure
762 * we have an exclusive reference.
763 */
764 sx_xlock(&allproc_lock);
765 LIST_REMOVE(p, p_list); /* off zombproc */
766 sx_xunlock(&allproc_lock);
767 LIST_REMOVE(p, p_sibling);
768 leavepgrp(p);
769 sx_xunlock(&proctree_lock);
770
771 /*
772 * As a side effect of this lock, we know that
773 * all other writes to this proc are visible now, so
774 * no more locking is needed for p.
775 */
776 PROC_LOCK(p);
777 p->p_xstat = 0; /* XXX: why? */
778 PROC_UNLOCK(p);
779 PROC_LOCK(q);
780 ruadd(&q->p_stats->p_cru, &q->p_crux, &p->p_ru,
781 &p->p_rux);
782 PROC_UNLOCK(q);
783
784 /*
785 * Decrement the count of procs running with this uid.
786 */
787 (void)chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0);
788
789 /*
790 * Free credentials, arguments, and sigacts.
791 */
792 crfree(p->p_ucred);
793 p->p_ucred = NULL;
794 pargs_drop(p->p_args);
795 p->p_args = NULL;
796 sigacts_free(p->p_sigacts);
797 p->p_sigacts = NULL;
798
799 /*
800 * Do any thread-system specific cleanups.
801 */
802 thread_wait(p);
803
804 /*
805 * Give vm and machine-dependent layer a chance
806 * to free anything that cpu_exit couldn't
807 * release while still running in process context.
808 */
809 vm_waitproc(p);
810 #ifdef MAC
811 mac_destroy_proc(p);
812 #endif
813 KASSERT(FIRST_THREAD_IN_PROC(p),
814 ("kern_wait: no residual thread!"));
815 uma_zfree(proc_zone, p);
816 sx_xlock(&allproc_lock);
817 nprocs--;
818 sx_xunlock(&allproc_lock);
819 return (0);
820 }
821 if ((p->p_flag & P_STOPPED_SIG) &&
822 (p->p_suspcount == p->p_numthreads) &&
823 (p->p_flag & P_WAITED) == 0 &&
824 (p->p_flag & P_TRACED || options & WUNTRACED)) {
825 PROC_SUNLOCK(p);
826 p->p_flag |= P_WAITED;
827 sx_xunlock(&proctree_lock);
828 td->td_retval[0] = p->p_pid;
829 if (status)
830 *status = W_STOPCODE(p->p_xstat);
831
832 PROC_LOCK(q);
833 sigqueue_take(p->p_ksi);
834 PROC_UNLOCK(q);
835 PROC_UNLOCK(p);
836
837 return (0);
838 }
839 PROC_SUNLOCK(p);
840 if (options & WCONTINUED && (p->p_flag & P_CONTINUED)) {
841 sx_xunlock(&proctree_lock);
842 td->td_retval[0] = p->p_pid;
843 p->p_flag &= ~P_CONTINUED;
844
845 PROC_LOCK(q);
846 sigqueue_take(p->p_ksi);
847 PROC_UNLOCK(q);
848 PROC_UNLOCK(p);
849
850 if (status)
851 *status = SIGCONT;
852 return (0);
853 }
854 PROC_UNLOCK(p);
855 }
856 if (nfound == 0) {
857 sx_xunlock(&proctree_lock);
858 return (ECHILD);
859 }
860 if (options & WNOHANG) {
861 sx_xunlock(&proctree_lock);
862 td->td_retval[0] = 0;
863 return (0);
864 }
865 PROC_LOCK(q);
866 sx_xunlock(&proctree_lock);
867 if (q->p_flag & P_STATCHILD) {
868 q->p_flag &= ~P_STATCHILD;
869 error = 0;
870 } else
871 error = msleep(q, &q->p_mtx, PWAIT | PCATCH, "wait", 0);
872 PROC_UNLOCK(q);
873 if (error)
874 return (error);
875 goto loop;
876 }
877
878 /*
879 * Make process 'parent' the new parent of process 'child'.
880 * Must be called with an exclusive hold of proctree lock.
881 */
882 void
883 proc_reparent(struct proc *child, struct proc *parent)
884 {
885
886 sx_assert(&proctree_lock, SX_XLOCKED);
887 PROC_LOCK_ASSERT(child, MA_OWNED);
888 if (child->p_pptr == parent)
889 return;
890
891 PROC_LOCK(child->p_pptr);
892 sigqueue_take(child->p_ksi);
893 PROC_UNLOCK(child->p_pptr);
894 LIST_REMOVE(child, p_sibling);
895 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
896 child->p_pptr = parent;
897 }
Cache object: 1c5f19801accaa987cd13b65f535e5f9
|