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: releng/8.3/sys/kern/kern_exit.c 220916 2011-04-21 09:01:43Z pluknet $");
39
40 #include "opt_compat.h"
41 #include "opt_kdtrace.h"
42 #include "opt_ktrace.h"
43
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/sysproto.h>
47 #include <sys/eventhandler.h>
48 #include <sys/kernel.h>
49 #include <sys/malloc.h>
50 #include <sys/lock.h>
51 #include <sys/mutex.h>
52 #include <sys/proc.h>
53 #include <sys/pioctl.h>
54 #include <sys/jail.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, exit);
94 SDT_PROBE_ARGTYPE(proc, kernel, , exit, 0, "int");
95
96 /* Hook for NFS teardown procedure. */
97 void (*nlminfo_release_p)(struct proc *p);
98
99 /*
100 * exit -- death of process.
101 */
102 void
103 sys_exit(struct thread *td, struct sys_exit_args *uap)
104 {
105
106 exit1(td, W_EXITCODE(uap->rval, 0));
107 /* NOTREACHED */
108 }
109
110 /*
111 * Exit: deallocate address space and other resources, change proc state to
112 * zombie, and unlink proc from allproc and parent's lists. Save exit status
113 * and rusage for wait(). Check for child processes and orphan them.
114 */
115 void
116 exit1(struct thread *td, int rv)
117 {
118 struct proc *p, *nq, *q;
119 struct vnode *vtmp;
120 struct vnode *ttyvp = NULL;
121 struct plimit *plim;
122 int locked;
123
124 mtx_assert(&Giant, MA_NOTOWNED);
125
126 p = td->td_proc;
127 /*
128 * XXX in case we're rebooting we just let init die in order to
129 * work around an unsolved stack overflow seen very late during
130 * shutdown on sparc64 when the gmirror worker process exists.
131 */
132 if (p == initproc && rebooting == 0) {
133 printf("init died (signal %d, exit %d)\n",
134 WTERMSIG(rv), WEXITSTATUS(rv));
135 panic("Going nowhere without my init!");
136 }
137
138 /*
139 * MUST abort all other threads before proceeding past here.
140 */
141 PROC_LOCK(p);
142 while (p->p_flag & P_HADTHREADS) {
143 /*
144 * First check if some other thread got here before us..
145 * if so, act apropriatly, (exit or suspend);
146 */
147 thread_suspend_check(0);
148
149 /*
150 * Kill off the other threads. This requires
151 * some co-operation from other parts of the kernel
152 * so it may not be instantaneous. With this state set
153 * any thread entering the kernel from userspace will
154 * thread_exit() in trap(). Any thread attempting to
155 * sleep will return immediately with EINTR or EWOULDBLOCK
156 * which will hopefully force them to back out to userland
157 * freeing resources as they go. Any thread attempting
158 * to return to userland will thread_exit() from userret().
159 * thread_exit() will unsuspend us when the last of the
160 * other threads exits.
161 * If there is already a thread singler after resumption,
162 * calling thread_single will fail; in that case, we just
163 * re-check all suspension request, the thread should
164 * either be suspended there or exit.
165 */
166 if (! thread_single(SINGLE_EXIT))
167 break;
168
169 /*
170 * All other activity in this process is now stopped.
171 * Threading support has been turned off.
172 */
173 }
174 KASSERT(p->p_numthreads == 1,
175 ("exit1: proc %p exiting with %d threads", p, p->p_numthreads));
176 /*
177 * Wakeup anyone in procfs' PIOCWAIT. They should have a hold
178 * on our vmspace, so we should block below until they have
179 * released their reference to us. Note that if they have
180 * requested S_EXIT stops we will block here until they ack
181 * via PIOCCONT.
182 */
183 _STOPEVENT(p, S_EXIT, rv);
184
185 /*
186 * Note that we are exiting and do another wakeup of anyone in
187 * PIOCWAIT in case they aren't listening for S_EXIT stops or
188 * decided to wait again after we told them we are exiting.
189 */
190 p->p_flag |= P_WEXIT;
191 wakeup(&p->p_stype);
192
193 /*
194 * Wait for any processes that have a hold on our vmspace to
195 * release their reference.
196 */
197 while (p->p_lock > 0)
198 msleep(&p->p_lock, &p->p_mtx, PWAIT, "exithold", 0);
199
200 p->p_xstat = rv; /* Let event handler change exit status */
201 PROC_UNLOCK(p);
202 /* Drain the limit callout while we don't have the proc locked */
203 callout_drain(&p->p_limco);
204
205 #ifdef AUDIT
206 /*
207 * The Sun BSM exit token contains two components: an exit status as
208 * passed to exit(), and a return value to indicate what sort of exit
209 * it was. The exit status is WEXITSTATUS(rv), but it's not clear
210 * what the return value is.
211 */
212 AUDIT_ARG_EXIT(WEXITSTATUS(rv), 0);
213 AUDIT_SYSCALL_EXIT(0, td);
214 #endif
215
216 /* Are we a task leader? */
217 if (p == p->p_leader) {
218 mtx_lock(&ppeers_lock);
219 q = p->p_peers;
220 while (q != NULL) {
221 PROC_LOCK(q);
222 psignal(q, SIGKILL);
223 PROC_UNLOCK(q);
224 q = q->p_peers;
225 }
226 while (p->p_peers != NULL)
227 msleep(p, &ppeers_lock, PWAIT, "exit1", 0);
228 mtx_unlock(&ppeers_lock);
229 }
230
231 /*
232 * Check if any loadable modules need anything done at process exit.
233 * E.g. SYSV IPC stuff
234 * XXX what if one of these generates an error?
235 */
236 EVENTHANDLER_INVOKE(process_exit, p);
237
238 /*
239 * If parent is waiting for us to exit or exec,
240 * P_PPWAIT is set; we will wakeup the parent below.
241 */
242 PROC_LOCK(p);
243 rv = p->p_xstat; /* Event handler could change exit status */
244 stopprofclock(p);
245 p->p_flag &= ~(P_TRACED | P_PPWAIT);
246
247 /*
248 * Stop the real interval timer. If the handler is currently
249 * executing, prevent it from rearming itself and let it finish.
250 */
251 if (timevalisset(&p->p_realtimer.it_value) &&
252 callout_stop(&p->p_itcallout) == 0) {
253 timevalclear(&p->p_realtimer.it_interval);
254 msleep(&p->p_itcallout, &p->p_mtx, PWAIT, "ritwait", 0);
255 KASSERT(!timevalisset(&p->p_realtimer.it_value),
256 ("realtime timer is still armed"));
257 }
258 PROC_UNLOCK(p);
259
260 /*
261 * Reset any sigio structures pointing to us as a result of
262 * F_SETOWN with our pid.
263 */
264 funsetownlst(&p->p_sigiolst);
265
266 /*
267 * If this process has an nlminfo data area (for lockd), release it
268 */
269 if (nlminfo_release_p != NULL && p->p_nlminfo != NULL)
270 (*nlminfo_release_p)(p);
271
272 /*
273 * Close open files and release open-file table.
274 * This may block!
275 */
276 fdfree(td);
277
278 /*
279 * If this thread tickled GEOM, we need to wait for the giggling to
280 * stop before we return to userland
281 */
282 if (td->td_pflags & TDP_GEOM)
283 g_waitidle();
284
285 /*
286 * Remove ourself from our leader's peer list and wake our leader.
287 */
288 mtx_lock(&ppeers_lock);
289 if (p->p_leader->p_peers) {
290 q = p->p_leader;
291 while (q->p_peers != p)
292 q = q->p_peers;
293 q->p_peers = p->p_peers;
294 wakeup(p->p_leader);
295 }
296 mtx_unlock(&ppeers_lock);
297
298 vmspace_exit(td);
299
300 sx_xlock(&proctree_lock);
301 if (SESS_LEADER(p)) {
302 struct session *sp = p->p_session;
303 struct tty *tp;
304
305 /*
306 * s_ttyp is not zero'd; we use this to indicate that
307 * the session once had a controlling terminal. (for
308 * logging and informational purposes)
309 */
310 SESS_LOCK(sp);
311 ttyvp = sp->s_ttyvp;
312 tp = sp->s_ttyp;
313 sp->s_ttyvp = NULL;
314 sp->s_leader = NULL;
315 SESS_UNLOCK(sp);
316
317 /*
318 * Signal foreground pgrp and revoke access to
319 * controlling terminal if it has not been revoked
320 * already.
321 *
322 * Because the TTY may have been revoked in the mean
323 * time and could already have a new session associated
324 * with it, make sure we don't send a SIGHUP to a
325 * foreground process group that does not belong to this
326 * session.
327 */
328
329 if (tp != NULL) {
330 tty_lock(tp);
331 if (tp->t_session == sp)
332 tty_signal_pgrp(tp, SIGHUP);
333 tty_unlock(tp);
334 }
335
336 if (ttyvp != NULL) {
337 sx_xunlock(&proctree_lock);
338 if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
339 VOP_REVOKE(ttyvp, REVOKEALL);
340 VOP_UNLOCK(ttyvp, 0);
341 }
342 sx_xlock(&proctree_lock);
343 }
344 }
345 fixjobc(p, p->p_pgrp, 0);
346 sx_xunlock(&proctree_lock);
347 (void)acct_process(td);
348
349 /* Release the TTY now we've unlocked everything. */
350 if (ttyvp != NULL)
351 vrele(ttyvp);
352 #ifdef KTRACE
353 ktrprocexit(td);
354 #endif
355 /*
356 * Release reference to text vnode
357 */
358 if ((vtmp = p->p_textvp) != NULL) {
359 p->p_textvp = NULL;
360 locked = VFS_LOCK_GIANT(vtmp->v_mount);
361 vrele(vtmp);
362 VFS_UNLOCK_GIANT(locked);
363 }
364
365 /*
366 * Release our limits structure.
367 */
368 PROC_LOCK(p);
369 plim = p->p_limit;
370 p->p_limit = NULL;
371 PROC_UNLOCK(p);
372 lim_free(plim);
373
374 /*
375 * Remove proc from allproc queue and pidhash chain.
376 * Place onto zombproc. Unlink from parent's child list.
377 */
378 sx_xlock(&allproc_lock);
379 LIST_REMOVE(p, p_list);
380 LIST_INSERT_HEAD(&zombproc, p, p_list);
381 LIST_REMOVE(p, p_hash);
382 sx_xunlock(&allproc_lock);
383
384 /*
385 * Call machine-dependent code to release any
386 * machine-dependent resources other than the address space.
387 * The address space is released by "vmspace_exitfree(p)" in
388 * vm_waitproc().
389 */
390 cpu_exit(td);
391
392 WITNESS_WARN(WARN_PANIC, NULL, "process (pid %d) exiting", p->p_pid);
393
394 /*
395 * Reparent all of our children to init.
396 */
397 sx_xlock(&proctree_lock);
398 q = LIST_FIRST(&p->p_children);
399 if (q != NULL) /* only need this if any child is S_ZOMB */
400 wakeup(initproc);
401 for (; q != NULL; q = nq) {
402 nq = LIST_NEXT(q, p_sibling);
403 PROC_LOCK(q);
404 proc_reparent(q, initproc);
405 q->p_sigparent = SIGCHLD;
406 /*
407 * Traced processes are killed
408 * since their existence means someone is screwing up.
409 */
410 if (q->p_flag & P_TRACED) {
411 struct thread *temp;
412
413 q->p_flag &= ~(P_TRACED | P_STOPPED_TRACE);
414 FOREACH_THREAD_IN_PROC(q, temp)
415 temp->td_dbgflags &= ~TDB_SUSPEND;
416 psignal(q, SIGKILL);
417 }
418 PROC_UNLOCK(q);
419 }
420
421 /* Save exit status. */
422 PROC_LOCK(p);
423 p->p_xthread = td;
424
425 /* Tell the prison that we are gone. */
426 prison_proc_free(p->p_ucred->cr_prison);
427
428 #ifdef KDTRACE_HOOKS
429 /*
430 * Tell the DTrace fasttrap provider about the exit if it
431 * has declared an interest.
432 */
433 if (dtrace_fasttrap_exit)
434 dtrace_fasttrap_exit(p);
435 #endif
436
437 /*
438 * Notify interested parties of our demise.
439 */
440 KNOTE_LOCKED(&p->p_klist, NOTE_EXIT);
441
442 #ifdef KDTRACE_HOOKS
443 int reason = CLD_EXITED;
444 if (WCOREDUMP(rv))
445 reason = CLD_DUMPED;
446 else if (WIFSIGNALED(rv))
447 reason = CLD_KILLED;
448 SDT_PROBE(proc, kernel, , exit, reason, 0, 0, 0, 0);
449 #endif
450
451 /*
452 * Just delete all entries in the p_klist. At this point we won't
453 * report any more events, and there are nasty race conditions that
454 * can beat us if we don't.
455 */
456 knlist_clear(&p->p_klist, 1);
457
458 /*
459 * Notify parent that we're gone. If parent has the PS_NOCLDWAIT
460 * flag set, or if the handler is set to SIG_IGN, notify process
461 * 1 instead (and hope it will handle this situation).
462 */
463 PROC_LOCK(p->p_pptr);
464 mtx_lock(&p->p_pptr->p_sigacts->ps_mtx);
465 if (p->p_pptr->p_sigacts->ps_flag & (PS_NOCLDWAIT | PS_CLDSIGIGN)) {
466 struct proc *pp;
467
468 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
469 pp = p->p_pptr;
470 PROC_UNLOCK(pp);
471 proc_reparent(p, initproc);
472 p->p_sigparent = SIGCHLD;
473 PROC_LOCK(p->p_pptr);
474
475 /*
476 * Notify parent, so in case he was wait(2)ing or
477 * executing waitpid(2) with our pid, he will
478 * continue.
479 */
480 wakeup(pp);
481 } else
482 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
483
484 if (p->p_pptr == initproc)
485 psignal(p->p_pptr, SIGCHLD);
486 else if (p->p_sigparent != 0) {
487 if (p->p_sigparent == SIGCHLD)
488 childproc_exited(p);
489 else /* LINUX thread */
490 psignal(p->p_pptr, p->p_sigparent);
491 }
492 sx_xunlock(&proctree_lock);
493
494 /*
495 * The state PRS_ZOMBIE prevents other proesses from sending
496 * signal to the process, to avoid memory leak, we free memory
497 * for signal queue at the time when the state is set.
498 */
499 sigqueue_flush(&p->p_sigqueue);
500 sigqueue_flush(&td->td_sigqueue);
501
502 /*
503 * We have to wait until after acquiring all locks before
504 * changing p_state. We need to avoid all possible context
505 * switches (including ones from blocking on a mutex) while
506 * marked as a zombie. We also have to set the zombie state
507 * before we release the parent process' proc lock to avoid
508 * a lost wakeup. So, we first call wakeup, then we grab the
509 * sched lock, update the state, and release the parent process'
510 * proc lock.
511 */
512 wakeup(p->p_pptr);
513 cv_broadcast(&p->p_pwait);
514 sched_exit(p->p_pptr, td);
515 PROC_SLOCK(p);
516 p->p_state = PRS_ZOMBIE;
517 PROC_UNLOCK(p->p_pptr);
518
519 /*
520 * Hopefully no one will try to deliver a signal to the process this
521 * late in the game.
522 */
523 knlist_destroy(&p->p_klist);
524
525 /*
526 * Save our children's rusage information in our exit rusage.
527 */
528 ruadd(&p->p_ru, &p->p_rux, &p->p_stats->p_cru, &p->p_crux);
529
530 /*
531 * Make sure the scheduler takes this thread out of its tables etc.
532 * This will also release this thread's reference to the ucred.
533 * Other thread parts to release include pcb bits and such.
534 */
535 thread_exit();
536 }
537
538
539 #ifndef _SYS_SYSPROTO_H_
540 struct abort2_args {
541 char *why;
542 int nargs;
543 void **args;
544 };
545 #endif
546
547 int
548 abort2(struct thread *td, struct abort2_args *uap)
549 {
550 struct proc *p = td->td_proc;
551 struct sbuf *sb;
552 void *uargs[16];
553 int error, i, sig;
554
555 /*
556 * Do it right now so we can log either proper call of abort2(), or
557 * note, that invalid argument was passed. 512 is big enough to
558 * handle 16 arguments' descriptions with additional comments.
559 */
560 sb = sbuf_new(NULL, NULL, 512, SBUF_FIXEDLEN);
561 sbuf_clear(sb);
562 sbuf_printf(sb, "%s(pid %d uid %d) aborted: ",
563 p->p_comm, p->p_pid, td->td_ucred->cr_uid);
564 /*
565 * Since we can't return from abort2(), send SIGKILL in cases, where
566 * abort2() was called improperly
567 */
568 sig = SIGKILL;
569 /* Prevent from DoSes from user-space. */
570 if (uap->nargs < 0 || uap->nargs > 16)
571 goto out;
572 if (uap->nargs > 0) {
573 if (uap->args == NULL)
574 goto out;
575 error = copyin(uap->args, uargs, uap->nargs * sizeof(void *));
576 if (error != 0)
577 goto out;
578 }
579 /*
580 * Limit size of 'reason' string to 128. Will fit even when
581 * maximal number of arguments was chosen to be logged.
582 */
583 if (uap->why != NULL) {
584 error = sbuf_copyin(sb, uap->why, 128);
585 if (error < 0)
586 goto out;
587 } else {
588 sbuf_printf(sb, "(null)");
589 }
590 if (uap->nargs > 0) {
591 sbuf_printf(sb, "(");
592 for (i = 0;i < uap->nargs; i++)
593 sbuf_printf(sb, "%s%p", i == 0 ? "" : ", ", uargs[i]);
594 sbuf_printf(sb, ")");
595 }
596 /*
597 * Final stage: arguments were proper, string has been
598 * successfully copied from userspace, and copying pointers
599 * from user-space succeed.
600 */
601 sig = SIGABRT;
602 out:
603 if (sig == SIGKILL) {
604 sbuf_trim(sb);
605 sbuf_printf(sb, " (Reason text inaccessible)");
606 }
607 sbuf_cat(sb, "\n");
608 sbuf_finish(sb);
609 log(LOG_INFO, "%s", sbuf_data(sb));
610 sbuf_delete(sb);
611 exit1(td, W_EXITCODE(0, sig));
612 return (0);
613 }
614
615
616 #ifdef COMPAT_43
617 /*
618 * The dirty work is handled by kern_wait().
619 */
620 int
621 owait(struct thread *td, struct owait_args *uap __unused)
622 {
623 int error, status;
624
625 error = kern_wait(td, WAIT_ANY, &status, 0, NULL);
626 if (error == 0)
627 td->td_retval[1] = status;
628 return (error);
629 }
630 #endif /* COMPAT_43 */
631
632 /*
633 * The dirty work is handled by kern_wait().
634 */
635 int
636 wait4(struct thread *td, struct wait_args *uap)
637 {
638 struct rusage ru, *rup;
639 int error, status;
640
641 if (uap->rusage != NULL)
642 rup = &ru;
643 else
644 rup = NULL;
645 error = kern_wait(td, uap->pid, &status, uap->options, rup);
646 if (uap->status != NULL && error == 0)
647 error = copyout(&status, uap->status, sizeof(status));
648 if (uap->rusage != NULL && error == 0)
649 error = copyout(&ru, uap->rusage, sizeof(struct rusage));
650 return (error);
651 }
652
653 /*
654 * Reap the remains of a zombie process and optionally return status and
655 * rusage. Asserts and will release both the proctree_lock and the process
656 * lock as part of its work.
657 */
658 static void
659 proc_reap(struct thread *td, struct proc *p, int *status, int options,
660 struct rusage *rusage)
661 {
662 struct proc *q, *t;
663
664 sx_assert(&proctree_lock, SA_XLOCKED);
665 PROC_LOCK_ASSERT(p, MA_OWNED);
666 PROC_SLOCK_ASSERT(p, MA_OWNED);
667 KASSERT(p->p_state == PRS_ZOMBIE, ("proc_reap: !PRS_ZOMBIE"));
668
669 q = td->td_proc;
670 if (rusage) {
671 *rusage = p->p_ru;
672 calcru(p, &rusage->ru_utime, &rusage->ru_stime);
673 }
674 PROC_SUNLOCK(p);
675 td->td_retval[0] = p->p_pid;
676 if (status)
677 *status = p->p_xstat; /* convert to int */
678 if (options & WNOWAIT) {
679 /*
680 * Only poll, returning the status. Caller does not wish to
681 * release the proc struct just yet.
682 */
683 PROC_UNLOCK(p);
684 sx_xunlock(&proctree_lock);
685 return;
686 }
687
688 PROC_LOCK(q);
689 sigqueue_take(p->p_ksi);
690 PROC_UNLOCK(q);
691 PROC_UNLOCK(p);
692
693 /*
694 * If we got the child via a ptrace 'attach', we need to give it back
695 * to the old parent.
696 */
697 if (p->p_oppid && (t = pfind(p->p_oppid)) != NULL) {
698 PROC_LOCK(p);
699 p->p_oppid = 0;
700 proc_reparent(p, t);
701 PROC_UNLOCK(p);
702 tdsignal(t, NULL, SIGCHLD, p->p_ksi);
703 wakeup(t);
704 cv_broadcast(&p->p_pwait);
705 PROC_UNLOCK(t);
706 sx_xunlock(&proctree_lock);
707 return;
708 }
709
710 /*
711 * Remove other references to this process to ensure we have an
712 * exclusive reference.
713 */
714 sx_xlock(&allproc_lock);
715 LIST_REMOVE(p, p_list); /* off zombproc */
716 sx_xunlock(&allproc_lock);
717 LIST_REMOVE(p, p_sibling);
718 leavepgrp(p);
719 sx_xunlock(&proctree_lock);
720
721 /*
722 * As a side effect of this lock, we know that all other writes to
723 * this proc are visible now, so no more locking is needed for p.
724 */
725 PROC_LOCK(p);
726 p->p_xstat = 0; /* XXX: why? */
727 PROC_UNLOCK(p);
728 PROC_LOCK(q);
729 ruadd(&q->p_stats->p_cru, &q->p_crux, &p->p_ru, &p->p_rux);
730 PROC_UNLOCK(q);
731
732 /*
733 * Decrement the count of procs running with this uid.
734 */
735 (void)chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0);
736
737 /*
738 * Free credentials, arguments, and sigacts.
739 */
740 crfree(p->p_ucred);
741 p->p_ucred = NULL;
742 pargs_drop(p->p_args);
743 p->p_args = NULL;
744 sigacts_free(p->p_sigacts);
745 p->p_sigacts = NULL;
746
747 /*
748 * Do any thread-system specific cleanups.
749 */
750 thread_wait(p);
751
752 /*
753 * Give vm and machine-dependent layer a chance to free anything that
754 * cpu_exit couldn't release while still running in process context.
755 */
756 vm_waitproc(p);
757 #ifdef MAC
758 mac_proc_destroy(p);
759 #endif
760 KASSERT(FIRST_THREAD_IN_PROC(p),
761 ("proc_reap: no residual thread!"));
762 uma_zfree(proc_zone, p);
763 sx_xlock(&allproc_lock);
764 nprocs--;
765 sx_xunlock(&allproc_lock);
766 }
767
768 int
769 kern_wait(struct thread *td, pid_t pid, int *status, int options,
770 struct rusage *rusage)
771 {
772 struct proc *p, *q;
773 int error, nfound;
774
775 AUDIT_ARG_PID(pid);
776 AUDIT_ARG_VALUE(options);
777
778 q = td->td_proc;
779 if (pid == 0) {
780 PROC_LOCK(q);
781 pid = -q->p_pgid;
782 PROC_UNLOCK(q);
783 }
784 if (options &~ (WUNTRACED|WNOHANG|WCONTINUED|WNOWAIT|WLINUXCLONE))
785 return (EINVAL);
786 loop:
787 if (q->p_flag & P_STATCHILD) {
788 PROC_LOCK(q);
789 q->p_flag &= ~P_STATCHILD;
790 PROC_UNLOCK(q);
791 }
792 nfound = 0;
793 sx_xlock(&proctree_lock);
794 LIST_FOREACH(p, &q->p_children, p_sibling) {
795 PROC_LOCK(p);
796 if (pid != WAIT_ANY &&
797 p->p_pid != pid && p->p_pgid != -pid) {
798 PROC_UNLOCK(p);
799 continue;
800 }
801 if (p_canwait(td, p)) {
802 PROC_UNLOCK(p);
803 continue;
804 }
805
806 /*
807 * This special case handles a kthread spawned by linux_clone
808 * (see linux_misc.c). The linux_wait4 and linux_waitpid
809 * functions need to be able to distinguish between waiting
810 * on a process and waiting on a thread. It is a thread if
811 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option
812 * signifies we want to wait for threads and not processes.
813 */
814 if ((p->p_sigparent != SIGCHLD) ^
815 ((options & WLINUXCLONE) != 0)) {
816 PROC_UNLOCK(p);
817 continue;
818 }
819
820 nfound++;
821 PROC_SLOCK(p);
822 if (p->p_state == PRS_ZOMBIE) {
823 proc_reap(td, p, status, options, rusage);
824 return (0);
825 }
826 if ((p->p_flag & P_STOPPED_SIG) &&
827 (p->p_suspcount == p->p_numthreads) &&
828 (p->p_flag & P_WAITED) == 0 &&
829 (p->p_flag & P_TRACED || options & WUNTRACED)) {
830 PROC_SUNLOCK(p);
831 p->p_flag |= P_WAITED;
832 sx_xunlock(&proctree_lock);
833 td->td_retval[0] = p->p_pid;
834 if (status)
835 *status = W_STOPCODE(p->p_xstat);
836
837 PROC_LOCK(q);
838 sigqueue_take(p->p_ksi);
839 PROC_UNLOCK(q);
840 PROC_UNLOCK(p);
841
842 return (0);
843 }
844 PROC_SUNLOCK(p);
845 if (options & WCONTINUED && (p->p_flag & P_CONTINUED)) {
846 sx_xunlock(&proctree_lock);
847 td->td_retval[0] = p->p_pid;
848 p->p_flag &= ~P_CONTINUED;
849
850 PROC_LOCK(q);
851 sigqueue_take(p->p_ksi);
852 PROC_UNLOCK(q);
853 PROC_UNLOCK(p);
854
855 if (status)
856 *status = SIGCONT;
857 return (0);
858 }
859 PROC_UNLOCK(p);
860 }
861 if (nfound == 0) {
862 sx_xunlock(&proctree_lock);
863 return (ECHILD);
864 }
865 if (options & WNOHANG) {
866 sx_xunlock(&proctree_lock);
867 td->td_retval[0] = 0;
868 return (0);
869 }
870 PROC_LOCK(q);
871 sx_xunlock(&proctree_lock);
872 if (q->p_flag & P_STATCHILD) {
873 q->p_flag &= ~P_STATCHILD;
874 error = 0;
875 } else
876 error = msleep(q, &q->p_mtx, PWAIT | PCATCH, "wait", 0);
877 PROC_UNLOCK(q);
878 if (error)
879 return (error);
880 goto loop;
881 }
882
883 /*
884 * Make process 'parent' the new parent of process 'child'.
885 * Must be called with an exclusive hold of proctree lock.
886 */
887 void
888 proc_reparent(struct proc *child, struct proc *parent)
889 {
890
891 sx_assert(&proctree_lock, SX_XLOCKED);
892 PROC_LOCK_ASSERT(child, MA_OWNED);
893 if (child->p_pptr == parent)
894 return;
895
896 PROC_LOCK(child->p_pptr);
897 sigqueue_take(child->p_ksi);
898 PROC_UNLOCK(child->p_pptr);
899 LIST_REMOVE(child, p_sibling);
900 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
901 child->p_pptr = parent;
902 }
Cache object: 2439f8a7484ef155767f87041aa7ec68
|