xref: /openbsd/sys/kern/kern_exit.c (revision 0c9ac863)
1 /*	$OpenBSD: kern_exit.c,v 1.240 2024/12/17 14:45:00 claudio Exp $	*/
2 /*	$NetBSD: kern_exit.c,v 1.39 1996/04/22 01:38:25 christos Exp $	*/
3 
4 /*
5  * Copyright (c) 1982, 1986, 1989, 1991, 1993
6  *	The Regents of the University of California.  All rights reserved.
7  * (c) UNIX System Laboratories, Inc.
8  * All or some portions of this file are derived from material licensed
9  * to the University of California by American Telephone and Telegraph
10  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
11  * the permission of UNIX System Laboratories, Inc.
12  *
13  * Redistribution and use in source and binary forms, with or without
14  * modification, are permitted provided that the following conditions
15  * are met:
16  * 1. Redistributions of source code must retain the above copyright
17  *    notice, this list of conditions and the following disclaimer.
18  * 2. Redistributions in binary form must reproduce the above copyright
19  *    notice, this list of conditions and the following disclaimer in the
20  *    documentation and/or other materials provided with the distribution.
21  * 3. Neither the name of the University nor the names of its contributors
22  *    may be used to endorse or promote products derived from this software
23  *    without specific prior written permission.
24  *
25  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
26  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35  * SUCH DAMAGE.
36  *
37  *	@(#)kern_exit.c	8.7 (Berkeley) 2/12/94
38  */
39 
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/proc.h>
43 #include <sys/time.h>
44 #include <sys/resource.h>
45 #include <sys/wait.h>
46 #include <sys/vnode.h>
47 #include <sys/malloc.h>
48 #include <sys/resourcevar.h>
49 #include <sys/ptrace.h>
50 #include <sys/acct.h>
51 #include <sys/filedesc.h>
52 #include <sys/signalvar.h>
53 #include <sys/sched.h>
54 #include <sys/ktrace.h>
55 #include <sys/pool.h>
56 #include <sys/mutex.h>
57 #ifdef SYSVSEM
58 #include <sys/sem.h>
59 #endif
60 #include <sys/witness.h>
61 
62 #include <sys/mount.h>
63 #include <sys/syscallargs.h>
64 
65 #include <uvm/uvm_extern.h>
66 
67 #include "kcov.h"
68 #if NKCOV > 0
69 #include <sys/kcov.h>
70 #endif
71 
72 void	proc_finish_wait(struct proc *, struct process *);
73 void	process_clear_orphan(struct process *);
74 void	process_zap(struct process *);
75 void	proc_free(struct proc *);
76 void	unveil_destroy(struct process *ps);
77 
78 /*
79  * exit --
80  *	Death of process.
81  */
82 int
sys_exit(struct proc * p,void * v,register_t * retval)83 sys_exit(struct proc *p, void *v, register_t *retval)
84 {
85 	struct sys_exit_args /* {
86 		syscallarg(int) rval;
87 	} */ *uap = v;
88 
89 	exit1(p, SCARG(uap, rval), 0, EXIT_NORMAL);
90 	/* NOTREACHED */
91 	return (0);
92 }
93 
94 int
sys___threxit(struct proc * p,void * v,register_t * retval)95 sys___threxit(struct proc *p, void *v, register_t *retval)
96 {
97 	struct sys___threxit_args /* {
98 		syscallarg(pid_t *) notdead;
99 	} */ *uap = v;
100 
101 	if (SCARG(uap, notdead) != NULL) {
102 		pid_t zero = 0;
103 		if (copyout(&zero, SCARG(uap, notdead), sizeof(zero)))
104 			psignal(p, SIGSEGV);
105 	}
106 	exit1(p, 0, 0, EXIT_THREAD);
107 
108 	return (0);
109 }
110 
111 /*
112  * Exit: deallocate address space and other resources, change proc state
113  * to zombie, and unlink proc from allproc and parent's lists.  Save exit
114  * status and rusage for wait().  Check for child processes and orphan them.
115  */
116 void
exit1(struct proc * p,int xexit,int xsig,int flags)117 exit1(struct proc *p, int xexit, int xsig, int flags)
118 {
119 	struct process *pr, *qr, *nqr;
120 	struct rusage *rup;
121 
122 	atomic_setbits_int(&p->p_flag, P_WEXIT);
123 
124 	pr = p->p_p;
125 
126 	/* single-threaded? */
127 	if (!P_HASSIBLING(p)) {
128 		flags = EXIT_NORMAL;
129 	} else {
130 		/* nope, multi-threaded */
131 		if (flags == EXIT_NORMAL)
132 			single_thread_set(p, SINGLE_EXIT);
133 	}
134 
135 	if (flags == EXIT_NORMAL && !(pr->ps_flags & PS_EXITING)) {
136 		if (pr->ps_pid == 1)
137 			panic("init died (signal %d, exit %d)", xsig, xexit);
138 
139 		atomic_setbits_int(&pr->ps_flags, PS_EXITING);
140 		pr->ps_xexit = xexit;
141 		pr->ps_xsig  = xsig;
142 
143 		/*
144 		 * If parent is waiting for us to exit or exec, PS_PPWAIT
145 		 * is set; we wake up the parent early to avoid deadlock.
146 		 */
147 		if (pr->ps_flags & PS_PPWAIT) {
148 			atomic_clearbits_int(&pr->ps_flags, PS_PPWAIT);
149 			atomic_clearbits_int(&pr->ps_pptr->ps_flags,
150 			    PS_ISPWAIT);
151 			wakeup(pr->ps_pptr);
152 		}
153 
154 		/* Wait for concurrent `allprocess' loops */
155 		refcnt_finalize(&pr->ps_refcnt, "psdtor");
156 	}
157 
158 	/* unlink ourselves from the active threads */
159 	mtx_enter(&pr->ps_mtx);
160 	TAILQ_REMOVE(&pr->ps_threads, p, p_thr_link);
161 	pr->ps_threadcnt--;
162 	pr->ps_exitcnt++;
163 
164 	/*
165 	 * if somebody else wants to take us to single threaded mode,
166 	 * count ourselves out.
167 	 */
168 	if (pr->ps_single) {
169 		if (--pr->ps_singlecnt == 0)
170 			wakeup(&pr->ps_singlecnt);
171 	}
172 
173 	/* proc is off ps_threads list so update accounting of process now */
174 	tuagg_add_runtime();
175 	tuagg_add_process(pr, p);
176 
177 	if ((p->p_flag & P_THREAD) == 0) {
178 		/* main thread gotta wait because it has the pid, et al */
179 		while (pr->ps_threadcnt + pr->ps_exitcnt > 1)
180 			msleep_nsec(&pr->ps_threads, &pr->ps_mtx, PWAIT,
181 			    "thrdeath", INFSLP);
182 	}
183 	mtx_leave(&pr->ps_mtx);
184 
185 	rup = pr->ps_ru;
186 	if (rup == NULL) {
187 		rup = pool_get(&rusage_pool, PR_WAITOK | PR_ZERO);
188 		if (pr->ps_ru == NULL) {
189 			pr->ps_ru = rup;
190 		} else {
191 			pool_put(&rusage_pool, rup);
192 			rup = pr->ps_ru;
193 		}
194 	}
195 	p->p_siglist = 0;
196 	if ((p->p_flag & P_THREAD) == 0)
197 		pr->ps_siglist = 0;
198 
199 	kqpoll_exit();
200 
201 #if NKCOV > 0
202 	kcov_exit(p);
203 #endif
204 
205 	if ((p->p_flag & P_THREAD) == 0) {
206 		if (pr->ps_flags & PS_PROFIL)
207 			stopprofclock(pr);
208 
209 		sigio_freelist(&pr->ps_sigiolst);
210 
211 		/* close open files and release open-file table */
212 		fdfree(p);
213 
214 		cancel_all_itimers();
215 
216 		timeout_del(&pr->ps_rucheck_to);
217 #ifdef SYSVSEM
218 		semexit(pr);
219 #endif
220 		killjobc(pr);
221 #ifdef ACCOUNTING
222 		acct_process(p);
223 #endif
224 
225 #ifdef KTRACE
226 		/* release trace file */
227 		if (pr->ps_tracevp)
228 			ktrcleartrace(pr);
229 #endif
230 
231 		unveil_destroy(pr);
232 
233 		free(pr->ps_pin.pn_pins, M_PINSYSCALL,
234 		    pr->ps_pin.pn_npins * sizeof(u_int));
235 		free(pr->ps_libcpin.pn_pins, M_PINSYSCALL,
236 		    pr->ps_libcpin.pn_npins * sizeof(u_int));
237 
238 		/*
239 		 * If parent has the SAS_NOCLDWAIT flag set, we're not
240 		 * going to become a zombie.
241 		 */
242 		if (pr->ps_pptr->ps_sigacts->ps_sigflags & SAS_NOCLDWAIT)
243 			atomic_setbits_int(&pr->ps_flags, PS_NOZOMBIE);
244 	}
245 
246 	p->p_fd = NULL;		/* zap the thread's copy */
247 
248 	/* Release the thread's read reference of resource limit structure. */
249 	if (p->p_limit != NULL) {
250 		struct plimit *limit;
251 
252 		limit = p->p_limit;
253 		p->p_limit = NULL;
254 		lim_free(limit);
255 	}
256 
257         /*
258 	 * Remove proc from pidhash chain and allproc so looking
259 	 * it up won't work.  We will put the proc on the
260 	 * deadproc list later (using the p_hash member), and
261 	 * wake up the reaper when we do.  If this is the last
262 	 * thread of a process that isn't PS_NOZOMBIE, we'll put
263 	 * the process on the zombprocess list below.
264 	 */
265 	/*
266 	 * NOTE: WE ARE NO LONGER ALLOWED TO SLEEP!
267 	 */
268 	p->p_stat = SDEAD;
269 
270 	LIST_REMOVE(p, p_hash);
271 	LIST_REMOVE(p, p_list);
272 
273 	if ((p->p_flag & P_THREAD) == 0) {
274 		LIST_REMOVE(pr, ps_hash);
275 		LIST_REMOVE(pr, ps_list);
276 
277 		if ((pr->ps_flags & PS_NOZOMBIE) == 0)
278 			LIST_INSERT_HEAD(&zombprocess, pr, ps_list);
279 		else {
280 			/*
281 			 * Not going to be a zombie, so it's now off all
282 			 * the lists scanned by ispidtaken(), so block
283 			 * fast reuse of the pid now.
284 			 */
285 			freepid(pr->ps_pid);
286 		}
287 
288 		/*
289 		 * Reparent children to their original parent, in case
290 		 * they were being traced, or to init(8).
291 		 */
292 		qr = LIST_FIRST(&pr->ps_children);
293 		if (qr)		/* only need this if any child is S_ZOMB */
294 			wakeup(initprocess);
295 		for (; qr != NULL; qr = nqr) {
296 			nqr = LIST_NEXT(qr, ps_sibling);
297 			/*
298 			 * Traced processes are killed since their
299 			 * existence means someone is screwing up.
300 			 */
301 			mtx_enter(&qr->ps_mtx);
302 			if (qr->ps_flags & PS_TRACED &&
303 			    !(qr->ps_flags & PS_EXITING)) {
304 				process_untrace(qr);
305 				mtx_leave(&qr->ps_mtx);
306 
307 				/*
308 				 * If single threading is active,
309 				 * direct the signal to the active
310 				 * thread to avoid deadlock.
311 				 */
312 				if (qr->ps_single)
313 					ptsignal(qr->ps_single, SIGKILL,
314 					    STHREAD);
315 				else
316 					prsignal(qr, SIGKILL);
317 			} else {
318 				process_reparent(qr, initprocess);
319 				mtx_leave(&qr->ps_mtx);
320 			}
321 		}
322 
323 		/*
324 		 * Make sure orphans won't remember the exiting process.
325 		 */
326 		while ((qr = LIST_FIRST(&pr->ps_orphans)) != NULL) {
327 			mtx_enter(&qr->ps_mtx);
328 			KASSERT(qr->ps_opptr == pr);
329 			qr->ps_opptr = NULL;
330 			process_clear_orphan(qr);
331 			mtx_leave(&qr->ps_mtx);
332 		}
333 	}
334 
335 	/* add thread's accumulated rusage into the process's total */
336 	ruadd(rup, &p->p_ru);
337 
338 	/*
339 	 * clear %cpu usage during swap
340 	 */
341 	p->p_pctcpu = 0;
342 
343 	if ((p->p_flag & P_THREAD) == 0) {
344 		/*
345 		 * Final thread has died, so add on our children's rusage
346 		 * and calculate the total times.
347 		 */
348 		calcru(&pr->ps_tu, &rup->ru_utime, &rup->ru_stime, NULL);
349 		ruadd(rup, &pr->ps_cru);
350 
351 		/*
352 		 * Notify parent that we're gone.  If we're not going to
353 		 * become a zombie, reparent to process 1 (init) so that
354 		 * we can wake our original parent to possibly unblock
355 		 * wait4() to return ECHILD.
356 		 */
357 		mtx_enter(&pr->ps_mtx);
358 		if (pr->ps_flags & PS_NOZOMBIE) {
359 			struct process *ppr = pr->ps_pptr;
360 			process_reparent(pr, initprocess);
361 			wakeup(ppr);
362 		}
363 		mtx_leave(&pr->ps_mtx);
364 	}
365 
366 	/* just a thread? check if last one standing. */
367 	if (p->p_flag & P_THREAD) {
368 		/* scheduler_wait_hook(pr->ps_mainproc, p); XXX */
369 		mtx_enter(&pr->ps_mtx);
370 		pr->ps_exitcnt--;
371 		if (pr->ps_threadcnt + pr->ps_exitcnt == 1)
372 			wakeup(&pr->ps_threads);
373 		mtx_leave(&pr->ps_mtx);
374 	}
375 
376 	/*
377 	 * Other substructures are freed from reaper and wait().
378 	 */
379 
380 	/*
381 	 * Finally, call machine-dependent code to switch to a new
382 	 * context (possibly the idle context).  Once we are no longer
383 	 * using the dead process's vmspace and stack, exit2() will be
384 	 * called to schedule those resources to be released by the
385 	 * reaper thread.
386 	 *
387 	 * Note that cpu_exit() will end with a call equivalent to
388 	 * cpu_switch(), finishing our execution (pun intended).
389 	 */
390 	cpu_exit(p);
391 	panic("cpu_exit returned");
392 }
393 
394 /*
395  * Locking of this proclist is special; it's accessed in a
396  * critical section of process exit, and thus locking it can't
397  * modify interrupt state.  We use a simple spin lock for this
398  * proclist.  We use the p_hash member to linkup to deadproc.
399  */
400 struct mutex deadproc_mutex =
401     MUTEX_INITIALIZER_FLAGS(IPL_NONE, "deadproc", MTX_NOWITNESS);
402 struct proclist deadproc = LIST_HEAD_INITIALIZER(deadproc);
403 
404 /*
405  * We are called from sched_idle() once it is safe to schedule the
406  * dead process's resources to be freed. So this is not allowed to sleep.
407  *
408  * We lock the deadproc list, place the proc on that list (using
409  * the p_hash member), and wake up the reaper.
410  */
411 void
exit2(struct proc * p)412 exit2(struct proc *p)
413 {
414 	/* account the remainder of time spent in exit1() */
415 	mtx_enter(&p->p_p->ps_mtx);
416 	tuagg_add_process(p->p_p, p);
417 	mtx_leave(&p->p_p->ps_mtx);
418 
419 	mtx_enter(&deadproc_mutex);
420 	LIST_INSERT_HEAD(&deadproc, p, p_hash);
421 	mtx_leave(&deadproc_mutex);
422 
423 	wakeup(&deadproc);
424 }
425 
426 void
proc_free(struct proc * p)427 proc_free(struct proc *p)
428 {
429 	crfree(p->p_ucred);
430 	pool_put(&proc_pool, p);
431 	atomic_dec_int(&nthreads);
432 }
433 
434 /*
435  * Process reaper.  This is run by a kernel thread to free the resources
436  * of a dead process.  Once the resources are free, the process becomes
437  * a zombie, and the parent is allowed to read the undead's status.
438  */
439 void
reaper(void * arg)440 reaper(void *arg)
441 {
442 	struct proc *p;
443 
444 	KERNEL_UNLOCK();
445 
446 	SCHED_ASSERT_UNLOCKED();
447 
448 	for (;;) {
449 		mtx_enter(&deadproc_mutex);
450 		while ((p = LIST_FIRST(&deadproc)) == NULL)
451 			msleep_nsec(&deadproc, &deadproc_mutex, PVM, "reaper",
452 			    INFSLP);
453 
454 		/* Remove us from the deadproc list. */
455 		LIST_REMOVE(p, p_hash);
456 		mtx_leave(&deadproc_mutex);
457 
458 		WITNESS_THREAD_EXIT(p);
459 
460 		/*
461 		 * Free the VM resources we're still holding on to.
462 		 * We must do this from a valid thread because doing
463 		 * so may block.
464 		 */
465 		uvm_uarea_free(p);
466 		p->p_vmspace = NULL;		/* zap the thread's copy */
467 
468 		if (p->p_flag & P_THREAD) {
469 			/* Just a thread */
470 			proc_free(p);
471 		} else {
472 			struct process *pr = p->p_p;
473 
474 			/* Release the rest of the process's vmspace */
475 			uvm_exit(pr);
476 
477 			KERNEL_LOCK();
478 			if ((pr->ps_flags & PS_NOZOMBIE) == 0) {
479 				/* Process is now a true zombie. */
480 				atomic_setbits_int(&pr->ps_flags, PS_ZOMBIE);
481 			}
482 
483 			/* Notify listeners of our demise and clean up. */
484 			knote_processexit(pr);
485 
486 			if (pr->ps_flags & PS_ZOMBIE) {
487 				/* Post SIGCHLD and wake up parent. */
488 				prsignal(pr->ps_pptr, SIGCHLD);
489 				wakeup(pr->ps_pptr);
490 			} else {
491 				/* No one will wait for us, just zap it. */
492 				process_zap(pr);
493 			}
494 			KERNEL_UNLOCK();
495 		}
496 	}
497 }
498 
499 int
dowait6(struct proc * q,idtype_t idtype,id_t id,int * statusp,int options,struct rusage * rusage,siginfo_t * info,register_t * retval)500 dowait6(struct proc *q, idtype_t idtype, id_t id, int *statusp, int options,
501     struct rusage *rusage, siginfo_t *info, register_t *retval)
502 {
503 	int nfound;
504 	struct process *pr;
505 	int error;
506 
507 	if (info != NULL)
508 		memset(info, 0, sizeof(*info));
509 
510 loop:
511 	nfound = 0;
512 	LIST_FOREACH(pr, &q->p_p->ps_children, ps_sibling) {
513 		if ((pr->ps_flags & PS_NOZOMBIE) ||
514 		    (idtype == P_PID && id != pr->ps_pid) ||
515 		    (idtype == P_PGID && id != pr->ps_pgid))
516 			continue;
517 
518 		nfound++;
519 		if ((options & WEXITED) && (pr->ps_flags & PS_ZOMBIE)) {
520 			*retval = pr->ps_pid;
521 			if (info != NULL) {
522 				info->si_pid = pr->ps_pid;
523 				info->si_uid = pr->ps_ucred->cr_uid;
524 				info->si_signo = SIGCHLD;
525 				if (pr->ps_xsig == 0) {
526 					info->si_code = CLD_EXITED;
527 					info->si_status = pr->ps_xexit;
528 				} else if (WCOREDUMP(pr->ps_xsig)) {
529 					info->si_code = CLD_DUMPED;
530 					info->si_status = _WSTATUS(pr->ps_xsig);
531 				} else {
532 					info->si_code = CLD_KILLED;
533 					info->si_status = _WSTATUS(pr->ps_xsig);
534 				}
535 			}
536 
537 			if (statusp != NULL)
538 				*statusp = W_EXITCODE(pr->ps_xexit,
539 				    pr->ps_xsig);
540 			if (rusage != NULL)
541 				memcpy(rusage, pr->ps_ru, sizeof(*rusage));
542 			if ((options & WNOWAIT) == 0)
543 				proc_finish_wait(q, pr);
544 			return (0);
545 		}
546 		if ((options & WTRAPPED) && (pr->ps_flags & PS_TRACED) &&
547 		    (pr->ps_flags & PS_WAITED) == 0 &&
548 		    (pr->ps_flags & PS_TRAPPED)) {
549 			if (single_thread_wait(pr, 0))
550 				goto loop;
551 
552 			if ((options & WNOWAIT) == 0)
553 				atomic_setbits_int(&pr->ps_flags, PS_WAITED);
554 
555 			*retval = pr->ps_pid;
556 			if (info != NULL) {
557 				info->si_pid = pr->ps_pid;
558 				info->si_uid = pr->ps_ucred->cr_uid;
559 				info->si_signo = SIGCHLD;
560 				info->si_code = CLD_TRAPPED;
561 				info->si_status = pr->ps_xsig;
562 			}
563 
564 			if (statusp != NULL)
565 				*statusp = W_STOPCODE(pr->ps_xsig);
566 			if (rusage != NULL)
567 				memset(rusage, 0, sizeof(*rusage));
568 			return (0);
569 		}
570 		if (((pr->ps_flags & PS_TRACED) || (options & WUNTRACED)) &&
571 		    (pr->ps_flags & PS_WAITED) == 0 &&
572 		    (pr->ps_flags & PS_STOPPED) &&
573 		    (pr->ps_flags & PS_TRAPPED) == 0) {
574 			if ((options & WNOWAIT) == 0)
575 				atomic_setbits_int(&pr->ps_flags, PS_WAITED);
576 
577 			*retval = pr->ps_pid;
578 			if (info != 0) {
579 				info->si_pid = pr->ps_pid;
580 				info->si_uid = pr->ps_ucred->cr_uid;
581 				info->si_signo = SIGCHLD;
582 				info->si_code = CLD_STOPPED;
583 				info->si_status = pr->ps_xsig;
584 			}
585 
586 			if (statusp != NULL)
587 				*statusp = W_STOPCODE(pr->ps_xsig);
588 			if (rusage != NULL)
589 				memset(rusage, 0, sizeof(*rusage));
590 			return (0);
591 		}
592 		if ((options & WCONTINUED) && (pr->ps_flags & PS_CONTINUED)) {
593 			if ((options & WNOWAIT) == 0)
594 				atomic_clearbits_int(&pr->ps_flags,
595 				    PS_CONTINUED);
596 
597 			*retval = pr->ps_pid;
598 			if (info != NULL) {
599 				info->si_pid = pr->ps_pid;
600 				info->si_uid = pr->ps_ucred->cr_uid;
601 				info->si_signo = SIGCHLD;
602 				info->si_code = CLD_CONTINUED;
603 				info->si_status = SIGCONT;
604 			}
605 
606 			if (statusp != NULL)
607 				*statusp = _WCONTINUED;
608 			if (rusage != NULL)
609 				memset(rusage, 0, sizeof(*rusage));
610 			return (0);
611 		}
612 	}
613 	/*
614 	 * Look in the orphans list too, to allow the parent to
615 	 * collect its child's exit status even if child is being
616 	 * debugged.
617 	 *
618 	 * Debugger detaches from the parent upon successful
619 	 * switch-over from parent to child.  At this point due to
620 	 * re-parenting the parent loses the child to debugger and a
621 	 * wait4(2) call would report that it has no children to wait
622 	 * for.  By maintaining a list of orphans we allow the parent
623 	 * to successfully wait until the child becomes a zombie.
624 	 */
625 	if (nfound == 0) {
626 		LIST_FOREACH(pr, &q->p_p->ps_orphans, ps_orphan) {
627 			if ((pr->ps_flags & PS_NOZOMBIE) ||
628 			    (idtype == P_PID && id != pr->ps_pid) ||
629 			    (idtype == P_PGID && id != pr->ps_pgid))
630 				continue;
631 			nfound++;
632 			break;
633 		}
634 	}
635 	if (nfound == 0)
636 		return (ECHILD);
637 	if (options & WNOHANG) {
638 		*retval = 0;
639 		return (0);
640 	}
641 	if ((error = tsleep_nsec(q->p_p, PWAIT | PCATCH, "wait", INFSLP)) != 0)
642 		return (error);
643 	goto loop;
644 }
645 
646 int
sys_wait4(struct proc * q,void * v,register_t * retval)647 sys_wait4(struct proc *q, void *v, register_t *retval)
648 {
649 	struct sys_wait4_args /* {
650 		syscallarg(pid_t) pid;
651 		syscallarg(int *) status;
652 		syscallarg(int) options;
653 		syscallarg(struct rusage *) rusage;
654 	} */ *uap = v;
655 	struct rusage ru;
656 	pid_t pid = SCARG(uap, pid);
657 	int options = SCARG(uap, options);
658 	int status, error;
659 	idtype_t idtype;
660 	id_t id;
661 
662 	if (SCARG(uap, options) &~ (WUNTRACED|WNOHANG|WCONTINUED))
663 		return (EINVAL);
664 	options |= WEXITED | WTRAPPED;
665 
666 	if (SCARG(uap, pid) == WAIT_MYPGRP) {
667 		idtype = P_PGID;
668 		id = q->p_p->ps_pgid;
669 	} else if (SCARG(uap, pid) == WAIT_ANY) {
670 		idtype = P_ALL;
671 		id = 0;
672 	} else if (pid < 0) {
673 		idtype = P_PGID;
674 		id = -pid;
675 	} else {
676 		idtype = P_PID;
677 		id = pid;
678 	}
679 
680 	error = dowait6(q, idtype, id,
681 	    SCARG(uap, status) ? &status : NULL, options,
682 	    SCARG(uap, rusage) ? &ru : NULL, NULL, retval);
683 	if (error == 0 && *retval > 0 && SCARG(uap, status)) {
684 		error = copyout(&status, SCARG(uap, status), sizeof(status));
685 	}
686 	if (error == 0 && *retval > 0 && SCARG(uap, rusage)) {
687 		error = copyout(&ru, SCARG(uap, rusage), sizeof(ru));
688 #ifdef KTRACE
689 		if (error == 0 && KTRPOINT(q, KTR_STRUCT))
690 			ktrrusage(q, &ru);
691 #endif
692 	}
693 	return (error);
694 }
695 
696 int
sys_waitid(struct proc * q,void * v,register_t * retval)697 sys_waitid(struct proc *q, void *v, register_t *retval)
698 {
699 	struct sys_waitid_args /* {
700 		syscallarg(idtype_t) idtype;
701 		syscallarg(id_t) id;
702 		syscallarg(siginfo_t) info;
703 		syscallarg(int) options;
704 	} */ *uap = v;
705 	siginfo_t info;
706 	idtype_t idtype = SCARG(uap, idtype);
707 	int options = SCARG(uap, options);
708 	int error;
709 
710 	if (options &~ (WSTOPPED|WCONTINUED|WEXITED|WTRAPPED|WNOHANG|WNOWAIT))
711 		return (EINVAL);
712 	if ((options & (WSTOPPED|WCONTINUED|WEXITED|WTRAPPED)) == 0)
713 		return (EINVAL);
714 	if (idtype != P_ALL && idtype != P_PID && idtype != P_PGID)
715 		return (EINVAL);
716 
717 	error = dowait6(q, idtype, SCARG(uap, id), NULL,
718 	    options, NULL, &info, retval);
719 	if (error == 0) {
720 		error = copyout(&info, SCARG(uap, info), sizeof(info));
721 #ifdef KTRACE
722 		if (error == 0 && KTRPOINT(q, KTR_STRUCT))
723 			ktrsiginfo(q, &info);
724 #endif
725 	}
726 	if (error == 0)
727 		*retval = 0;
728 	return (error);
729 }
730 
731 void
proc_finish_wait(struct proc * waiter,struct process * pr)732 proc_finish_wait(struct proc *waiter, struct process *pr)
733 {
734 	struct process *tr;
735 	struct rusage *rup;
736 
737 	/*
738 	 * If we got the child via a ptrace 'attach',
739 	 * we need to give it back to the old parent.
740 	 */
741 	mtx_enter(&pr->ps_mtx);
742 	if (pr->ps_opptr != NULL && (pr->ps_opptr != pr->ps_pptr)) {
743 		tr = pr->ps_opptr;
744 		pr->ps_opptr = NULL;
745 		atomic_clearbits_int(&pr->ps_flags, PS_TRACED);
746 		process_reparent(pr, tr);
747 		mtx_leave(&pr->ps_mtx);
748 		prsignal(tr, SIGCHLD);
749 		wakeup(tr);
750 	} else {
751 		mtx_leave(&pr->ps_mtx);
752 		scheduler_wait_hook(waiter, pr->ps_mainproc);
753 		rup = &waiter->p_p->ps_cru;
754 		ruadd(rup, pr->ps_ru);
755 		LIST_REMOVE(pr, ps_list);	/* off zombprocess */
756 		freepid(pr->ps_pid);
757 		process_zap(pr);
758 	}
759 }
760 
761 /*
762  * give process back to original parent or init(8)
763  */
764 void
process_untrace(struct process * pr)765 process_untrace(struct process *pr)
766 {
767 	struct process *ppr = NULL;
768 
769 	KASSERT(pr->ps_flags & PS_TRACED);
770 	MUTEX_ASSERT_LOCKED(&pr->ps_mtx);
771 
772 	if (pr->ps_opptr != NULL &&
773 	    (pr->ps_opptr != pr->ps_pptr))
774 		ppr = pr->ps_opptr;
775 
776 	/* not being traced any more */
777 	pr->ps_opptr = NULL;
778 	atomic_clearbits_int(&pr->ps_flags, PS_TRACED);
779 	process_reparent(pr, ppr ? ppr : initprocess);
780 }
781 
782 void
process_clear_orphan(struct process * pr)783 process_clear_orphan(struct process *pr)
784 {
785 	if (pr->ps_flags & PS_ORPHAN) {
786 		LIST_REMOVE(pr, ps_orphan);
787 		atomic_clearbits_int(&pr->ps_flags, PS_ORPHAN);
788 	}
789 }
790 
791 /*
792  * make process 'parent' the new parent of process 'child'.
793  */
794 void
process_reparent(struct process * child,struct process * parent)795 process_reparent(struct process *child, struct process *parent)
796 {
797 
798 	if (child->ps_pptr == parent)
799 		return;
800 
801 	KASSERT(child->ps_opptr == NULL ||
802 		child->ps_opptr == child->ps_pptr);
803 
804 	LIST_REMOVE(child, ps_sibling);
805 	LIST_INSERT_HEAD(&parent->ps_children, child, ps_sibling);
806 
807 	process_clear_orphan(child);
808 	if (child->ps_flags & PS_TRACED) {
809 		atomic_setbits_int(&child->ps_flags, PS_ORPHAN);
810 		LIST_INSERT_HEAD(&child->ps_pptr->ps_orphans, child, ps_orphan);
811 	}
812 
813 	MUTEX_ASSERT_LOCKED(&child->ps_mtx);
814 	child->ps_pptr = parent;
815 	child->ps_ppid = parent->ps_pid;
816 }
817 
818 void
process_zap(struct process * pr)819 process_zap(struct process *pr)
820 {
821 	struct vnode *otvp;
822 	struct proc *p = pr->ps_mainproc;
823 
824 	/*
825 	 * Finally finished with old proc entry.
826 	 * Unlink it from its process group and free it.
827 	 */
828 	leavepgrp(pr);
829 	LIST_REMOVE(pr, ps_sibling);
830 	process_clear_orphan(pr);
831 
832 	/*
833 	 * Decrement the count of procs running with this uid.
834 	 */
835 	(void)chgproccnt(pr->ps_ucred->cr_ruid, -1);
836 
837 	/*
838 	 * Release reference to text vnode
839 	 */
840 	otvp = pr->ps_textvp;
841 	pr->ps_textvp = NULL;
842 	if (otvp)
843 		vrele(otvp);
844 
845 	KASSERT(pr->ps_threadcnt == 0);
846 	KASSERT(pr->ps_exitcnt == 1);
847 	if (pr->ps_ptstat != NULL)
848 		free(pr->ps_ptstat, M_SUBPROC, sizeof(*pr->ps_ptstat));
849 	pool_put(&rusage_pool, pr->ps_ru);
850 	KASSERT(TAILQ_EMPTY(&pr->ps_threads));
851 	sigactsfree(pr->ps_sigacts);
852 	lim_free(pr->ps_limit);
853 	crfree(pr->ps_ucred);
854 	pool_put(&process_pool, pr);
855 	nprocesses--;
856 
857 	proc_free(p);
858 }
859