xref: /openbsd-src/sys/kern/kern_fork.c (revision 0747e3d20bf138958765ed50a0a2854bbcb6704f)
1 /*	$OpenBSD: kern_fork.c,v 1.261 2024/08/06 08:44:54 claudio Exp $	*/
2 /*	$NetBSD: kern_fork.c,v 1.29 1996/02/09 18:59:34 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_fork.c	8.6 (Berkeley) 4/8/94
38  */
39 
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/filedesc.h>
43 #include <sys/malloc.h>
44 #include <sys/mount.h>
45 #include <sys/proc.h>
46 #include <sys/resourcevar.h>
47 #include <sys/signalvar.h>
48 #include <sys/vnode.h>
49 #include <sys/vmmeter.h>
50 #include <sys/acct.h>
51 #include <sys/ktrace.h>
52 #include <sys/sched.h>
53 #include <sys/smr.h>
54 #include <sys/sysctl.h>
55 #include <sys/pool.h>
56 #include <sys/mman.h>
57 #include <sys/ptrace.h>
58 #include <sys/atomic.h>
59 #include <sys/unistd.h>
60 #include <sys/tracepoint.h>
61 
62 #include <sys/syscallargs.h>
63 
64 #include <uvm/uvm.h>
65 #include <machine/tcb.h>
66 
67 int	nprocesses = 1;		/* process 0 */
68 int	nthreads = 1;		/* proc 0 */
69 struct	forkstat forkstat;
70 
71 void fork_return(void *);
72 pid_t alloctid(void);
73 pid_t allocpid(void);
74 int ispidtaken(pid_t);
75 
76 void unveil_copy(struct process *parent, struct process *child);
77 
78 struct proc *thread_new(struct proc *_parent, vaddr_t _uaddr);
79 struct process *process_new(struct proc *, struct process *, int);
80 int fork_check_maxthread(uid_t _uid);
81 
82 void
83 fork_return(void *arg)
84 {
85 	struct proc *p = (struct proc *)arg;
86 
87 	if (p->p_p->ps_flags & PS_TRACED)
88 		psignal(p, SIGTRAP);
89 
90 	child_return(p);
91 }
92 
93 int
94 sys_fork(struct proc *p, void *v, register_t *retval)
95 {
96 	void (*func)(void *) = child_return;
97 	int flags;
98 
99 	flags = FORK_FORK;
100 	if (p->p_p->ps_ptmask & PTRACE_FORK) {
101 		flags |= FORK_PTRACE;
102 		func = fork_return;
103 	}
104 	return fork1(p, flags, func, NULL, retval, NULL);
105 }
106 
107 int
108 sys_vfork(struct proc *p, void *v, register_t *retval)
109 {
110 	return fork1(p, FORK_VFORK|FORK_PPWAIT, child_return, NULL,
111 	    retval, NULL);
112 }
113 
114 int
115 sys___tfork(struct proc *p, void *v, register_t *retval)
116 {
117 	struct sys___tfork_args /* {
118 		syscallarg(const struct __tfork) *param;
119 		syscallarg(size_t) psize;
120 	} */ *uap = v;
121 	size_t psize = SCARG(uap, psize);
122 	struct __tfork param = { 0 };
123 	int error;
124 
125 	if (psize == 0 || psize > sizeof(param))
126 		return EINVAL;
127 	if ((error = copyin(SCARG(uap, param), &param, psize)))
128 		return error;
129 #ifdef KTRACE
130 	if (KTRPOINT(p, KTR_STRUCT))
131 		ktrstruct(p, "tfork", &param, sizeof(param));
132 #endif
133 #ifdef TCB_INVALID
134 	if (TCB_INVALID(param.tf_tcb))
135 		return EINVAL;
136 #endif /* TCB_INVALID */
137 
138 	return thread_fork(p, param.tf_stack, param.tf_tcb, param.tf_tid,
139 	    retval);
140 }
141 
142 /*
143  * Allocate and initialize a thread (proc) structure, given the parent thread.
144  */
145 struct proc *
146 thread_new(struct proc *parent, vaddr_t uaddr)
147 {
148 	struct proc *p;
149 
150 	p = pool_get(&proc_pool, PR_WAITOK);
151 	p->p_stat = SIDL;			/* protect against others */
152 	p->p_runpri = 0;
153 	p->p_flag = 0;
154 
155 	/*
156 	 * Make a proc table entry for the new process.
157 	 * Start by zeroing the section of proc that is zero-initialized,
158 	 * then copy the section that is copied directly from the parent.
159 	 */
160 	memset(&p->p_startzero, 0,
161 	    (caddr_t)&p->p_endzero - (caddr_t)&p->p_startzero);
162 	memcpy(&p->p_startcopy, &parent->p_startcopy,
163 	    (caddr_t)&p->p_endcopy - (caddr_t)&p->p_startcopy);
164 	crhold(p->p_ucred);
165 	p->p_addr = (struct user *)uaddr;
166 
167 	/*
168 	 * Initialize the timeouts.
169 	 */
170 	timeout_set(&p->p_sleep_to, endtsleep, p);
171 
172 	return p;
173 }
174 
175 /*
176  * Initialize common bits of a process structure, given the initial thread.
177  */
178 void
179 process_initialize(struct process *pr, struct proc *p)
180 {
181 	/* initialize the thread links */
182 	pr->ps_mainproc = p;
183 	TAILQ_INIT(&pr->ps_threads);
184 	TAILQ_INSERT_TAIL(&pr->ps_threads, p, p_thr_link);
185 	pr->ps_threadcnt = 1;
186 	p->p_p = pr;
187 
188 	/* give the process the same creds as the initial thread */
189 	pr->ps_ucred = p->p_ucred;
190 	crhold(pr->ps_ucred);
191 	/* new thread and new process */
192 	KASSERT(p->p_ucred->cr_refcnt.r_refs >= 2);
193 
194 	LIST_INIT(&pr->ps_children);
195 	LIST_INIT(&pr->ps_orphans);
196 	LIST_INIT(&pr->ps_ftlist);
197 	LIST_INIT(&pr->ps_sigiolst);
198 	TAILQ_INIT(&pr->ps_tslpqueue);
199 
200 	rw_init(&pr->ps_lock, "pslock");
201 	mtx_init(&pr->ps_mtx, IPL_HIGH);
202 	klist_init_mutex(&pr->ps_klist, &pr->ps_mtx);
203 
204 	timeout_set_flags(&pr->ps_realit_to, realitexpire, pr,
205 	    KCLOCK_UPTIME, 0);
206 	timeout_set(&pr->ps_rucheck_to, rucheck, pr);
207 }
208 
209 
210 /*
211  * Allocate and initialize a new process.
212  */
213 struct process *
214 process_new(struct proc *p, struct process *parent, int flags)
215 {
216 	struct process *pr;
217 
218 	pr = pool_get(&process_pool, PR_WAITOK);
219 
220 	/*
221 	 * Make a process structure for the new process.
222 	 * Start by zeroing the section of proc that is zero-initialized,
223 	 * then copy the section that is copied directly from the parent.
224 	 */
225 	memset(&pr->ps_startzero, 0,
226 	    (caddr_t)&pr->ps_endzero - (caddr_t)&pr->ps_startzero);
227 	memcpy(&pr->ps_startcopy, &parent->ps_startcopy,
228 	    (caddr_t)&pr->ps_endcopy - (caddr_t)&pr->ps_startcopy);
229 
230 	process_initialize(pr, p);
231 	pr->ps_pid = allocpid();
232 	lim_fork(parent, pr);
233 
234 	/* post-copy fixups */
235 	pr->ps_pptr = parent;
236 	pr->ps_ppid = parent->ps_pid;
237 
238 	/* bump references to the text vnode (for sysctl) */
239 	pr->ps_textvp = parent->ps_textvp;
240 	if (pr->ps_textvp)
241 		vref(pr->ps_textvp);
242 
243 	/* copy unveil if unveil is active */
244 	unveil_copy(parent, pr);
245 
246 	pr->ps_flags = parent->ps_flags &
247 	    (PS_SUGID | PS_SUGIDEXEC | PS_PLEDGE | PS_EXECPLEDGE |
248 	    PS_WXNEEDED | PS_CHROOT);
249 	if (parent->ps_session->s_ttyvp != NULL)
250 		pr->ps_flags |= parent->ps_flags & PS_CONTROLT;
251 
252 	if (parent->ps_pin.pn_pins) {
253 		pr->ps_pin.pn_pins = mallocarray(parent->ps_pin.pn_npins,
254 		    sizeof(u_int), M_PINSYSCALL, M_WAITOK);
255 		memcpy(pr->ps_pin.pn_pins, parent->ps_pin.pn_pins,
256 		    parent->ps_pin.pn_npins * sizeof(u_int));
257 		pr->ps_flags |= PS_PIN;
258 	}
259 	if (parent->ps_libcpin.pn_pins) {
260 		pr->ps_libcpin.pn_pins = mallocarray(parent->ps_libcpin.pn_npins,
261 		    sizeof(u_int), M_PINSYSCALL, M_WAITOK);
262 		memcpy(pr->ps_libcpin.pn_pins, parent->ps_libcpin.pn_pins,
263 		    parent->ps_libcpin.pn_npins * sizeof(u_int));
264 		pr->ps_flags |= PS_LIBCPIN;
265 	}
266 
267 	/*
268 	 * Duplicate sub-structures as needed.
269 	 * Increase reference counts on shared objects.
270 	 */
271 	if (flags & FORK_SHAREFILES)
272 		pr->ps_fd = fdshare(parent);
273 	else
274 		pr->ps_fd = fdcopy(parent);
275 	pr->ps_sigacts = sigactsinit(parent);
276 	if (flags & FORK_SHAREVM)
277 		pr->ps_vmspace = uvmspace_share(parent);
278 	else
279 		pr->ps_vmspace = uvmspace_fork(parent);
280 
281 	if (parent->ps_flags & PS_PROFIL)
282 		startprofclock(pr);
283 	if (flags & FORK_PTRACE)
284 		pr->ps_flags |= parent->ps_flags & PS_TRACED;
285 	if (flags & FORK_NOZOMBIE)
286 		pr->ps_flags |= PS_NOZOMBIE;
287 	if (flags & FORK_SYSTEM)
288 		pr->ps_flags |= PS_SYSTEM;
289 
290 	/* mark as embryo to protect against others */
291 	pr->ps_flags |= PS_EMBRYO;
292 
293 	/* Force visibility of all of the above changes */
294 	membar_producer();
295 
296 	/* it's sufficiently inited to be globally visible */
297 	LIST_INSERT_HEAD(&allprocess, pr, ps_list);
298 
299 	return pr;
300 }
301 
302 /* print the 'table full' message once per 10 seconds */
303 struct timeval fork_tfmrate = { 10, 0 };
304 
305 int
306 fork_check_maxthread(uid_t uid)
307 {
308 	/*
309 	 * Although process entries are dynamically created, we still keep
310 	 * a global limit on the maximum number we will create. We reserve
311 	 * the last 5 processes to root. The variable nprocesses is the
312 	 * current number of processes, maxprocess is the limit.  Similar
313 	 * rules for threads (struct proc): we reserve the last 5 to root;
314 	 * the variable nthreads is the current number of procs, maxthread is
315 	 * the limit.
316 	 */
317 	if ((nthreads >= maxthread - 5 && uid != 0) || nthreads >= maxthread) {
318 		static struct timeval lasttfm;
319 
320 		if (ratecheck(&lasttfm, &fork_tfmrate))
321 			tablefull("thread");
322 		return EAGAIN;
323 	}
324 	nthreads++;
325 
326 	return 0;
327 }
328 
329 static inline void
330 fork_thread_start(struct proc *p, struct proc *parent, int flags)
331 {
332 	struct cpu_info *ci;
333 
334 	SCHED_LOCK();
335 	ci = sched_choosecpu_fork(parent, flags);
336 	TRACEPOINT(sched, fork, p->p_tid + THREAD_PID_OFFSET,
337 	    p->p_p->ps_pid, CPU_INFO_UNIT(ci));
338 	setrunqueue(ci, p, p->p_usrpri);
339 	SCHED_UNLOCK();
340 }
341 
342 int
343 fork1(struct proc *curp, int flags, void (*func)(void *), void *arg,
344     register_t *retval, struct proc **rnewprocp)
345 {
346 	struct process *curpr = curp->p_p;
347 	struct process *pr;
348 	struct proc *p;
349 	uid_t uid = curp->p_ucred->cr_ruid;
350 	struct vmspace *vm;
351 	int count;
352 	vaddr_t uaddr;
353 	int error;
354 	struct  ptrace_state *newptstat = NULL;
355 
356 	KASSERT((flags & ~(FORK_FORK | FORK_VFORK | FORK_PPWAIT | FORK_PTRACE
357 	    | FORK_IDLE | FORK_SHAREVM | FORK_SHAREFILES | FORK_NOZOMBIE
358 	    | FORK_SYSTEM)) == 0);
359 	KASSERT(func != NULL);
360 
361 	if ((error = fork_check_maxthread(uid)))
362 		return error;
363 
364 	if ((nprocesses >= maxprocess - 5 && uid != 0) ||
365 	    nprocesses >= maxprocess) {
366 		static struct timeval lasttfm;
367 
368 		if (ratecheck(&lasttfm, &fork_tfmrate))
369 			tablefull("process");
370 		nthreads--;
371 		return EAGAIN;
372 	}
373 	nprocesses++;
374 
375 	/*
376 	 * Increment the count of processes running with this uid.
377 	 * Don't allow a nonprivileged user to exceed their current limit.
378 	 */
379 	count = chgproccnt(uid, 1);
380 	if (uid != 0 && count > lim_cur(RLIMIT_NPROC)) {
381 		(void)chgproccnt(uid, -1);
382 		nprocesses--;
383 		nthreads--;
384 		return EAGAIN;
385 	}
386 
387 	uaddr = uvm_uarea_alloc();
388 	if (uaddr == 0) {
389 		(void)chgproccnt(uid, -1);
390 		nprocesses--;
391 		nthreads--;
392 		return (ENOMEM);
393 	}
394 
395 	/*
396 	 * From now on, we're committed to the fork and cannot fail.
397 	 */
398 	p = thread_new(curp, uaddr);
399 	pr = process_new(p, curpr, flags);
400 
401 	p->p_fd		= pr->ps_fd;
402 	p->p_vmspace	= pr->ps_vmspace;
403 	if (pr->ps_flags & PS_SYSTEM)
404 		atomic_setbits_int(&p->p_flag, P_SYSTEM);
405 
406 	if (flags & FORK_PPWAIT) {
407 		atomic_setbits_int(&pr->ps_flags, PS_PPWAIT);
408 		atomic_setbits_int(&curpr->ps_flags, PS_ISPWAIT);
409 	}
410 
411 #ifdef KTRACE
412 	/*
413 	 * Copy traceflag and tracefile if enabled.
414 	 * If not inherited, these were zeroed above.
415 	 */
416 	if (curpr->ps_traceflag & KTRFAC_INHERIT)
417 		ktrsettrace(pr, curpr->ps_traceflag, curpr->ps_tracevp,
418 		    curpr->ps_tracecred);
419 #endif
420 
421 	/*
422 	 * Finish creating the child thread.  cpu_fork() will copy
423 	 * and update the pcb and make the child ready to run.  If
424 	 * this is a normal user fork, the child will exit directly
425 	 * to user mode via child_return() on its first time slice
426 	 * and will not return here.  If this is a kernel thread,
427 	 * the specified entry point will be executed.
428 	 */
429 	cpu_fork(curp, p, NULL, NULL, func, arg ? arg : p);
430 
431 	vm = pr->ps_vmspace;
432 
433 	if (flags & FORK_FORK) {
434 		forkstat.cntfork++;
435 		forkstat.sizfork += vm->vm_dsize + vm->vm_ssize;
436 	} else if (flags & FORK_VFORK) {
437 		forkstat.cntvfork++;
438 		forkstat.sizvfork += vm->vm_dsize + vm->vm_ssize;
439 	} else {
440 		forkstat.cntkthread++;
441 	}
442 
443 	if (pr->ps_flags & PS_TRACED && flags & FORK_FORK)
444 		newptstat = malloc(sizeof(*newptstat), M_SUBPROC, M_WAITOK);
445 
446 	p->p_tid = alloctid();
447 
448 	LIST_INSERT_HEAD(&allproc, p, p_list);
449 	LIST_INSERT_HEAD(TIDHASH(p->p_tid), p, p_hash);
450 	LIST_INSERT_HEAD(PIDHASH(pr->ps_pid), pr, ps_hash);
451 	LIST_INSERT_AFTER(curpr, pr, ps_pglist);
452 	LIST_INSERT_HEAD(&curpr->ps_children, pr, ps_sibling);
453 
454 	if (pr->ps_flags & PS_TRACED) {
455 		pr->ps_oppid = curpr->ps_pid;
456 		process_reparent(pr, curpr->ps_pptr);
457 
458 		/*
459 		 * Set ptrace status.
460 		 */
461 		if (newptstat != NULL) {
462 			pr->ps_ptstat = newptstat;
463 			newptstat = NULL;
464 			curpr->ps_ptstat->pe_report_event = PTRACE_FORK;
465 			pr->ps_ptstat->pe_report_event = PTRACE_FORK;
466 			curpr->ps_ptstat->pe_other_pid = pr->ps_pid;
467 			pr->ps_ptstat->pe_other_pid = curpr->ps_pid;
468 		}
469 	}
470 
471 	/*
472 	 * For new processes, set accounting bits and mark as complete.
473 	 */
474 	nanouptime(&pr->ps_start);
475 	pr->ps_acflag = AFORK;
476 	atomic_clearbits_int(&pr->ps_flags, PS_EMBRYO);
477 
478 	if ((flags & FORK_IDLE) == 0)
479 		fork_thread_start(p, curp, flags);
480 	else
481 		p->p_cpu = arg;
482 
483 	free(newptstat, M_SUBPROC, sizeof(*newptstat));
484 
485 	/*
486 	 * Notify any interested parties about the new process.
487 	 */
488 	knote_processfork(curpr, pr->ps_pid);
489 
490 	/*
491 	 * Update stats now that we know the fork was successful.
492 	 */
493 	uvmexp.forks++;
494 	if (flags & FORK_PPWAIT)
495 		uvmexp.forks_ppwait++;
496 	if (flags & FORK_SHAREVM)
497 		uvmexp.forks_sharevm++;
498 
499 	/*
500 	 * Pass a pointer to the new process to the caller.
501 	 */
502 	if (rnewprocp != NULL)
503 		*rnewprocp = p;
504 
505 	/*
506 	 * Preserve synchronization semantics of vfork.  If waiting for
507 	 * child to exec or exit, set PS_PPWAIT on child and PS_ISPWAIT
508 	 * on ourselves, and sleep on our process for the latter flag
509 	 * to go away.
510 	 * XXX Need to stop other rthreads in the parent
511 	 */
512 	if (flags & FORK_PPWAIT)
513 		while (curpr->ps_flags & PS_ISPWAIT)
514 			tsleep_nsec(curpr, PWAIT, "ppwait", INFSLP);
515 
516 	/*
517 	 * If we're tracing the child, alert the parent too.
518 	 */
519 	if ((flags & FORK_PTRACE) && (curpr->ps_flags & PS_TRACED))
520 		psignal(curp, SIGTRAP);
521 
522 	/*
523 	 * Return child pid to parent process
524 	 */
525 	if (retval != NULL)
526 		*retval = pr->ps_pid;
527 	return (0);
528 }
529 
530 int
531 thread_fork(struct proc *curp, void *stack, void *tcb, pid_t *tidptr,
532     register_t *retval)
533 {
534 	struct process *pr = curp->p_p;
535 	struct proc *p;
536 	pid_t tid;
537 	vaddr_t uaddr;
538 	int error;
539 
540 	if (stack == NULL)
541 		return EINVAL;
542 
543 	if ((error = fork_check_maxthread(curp->p_ucred->cr_ruid)))
544 		return error;
545 
546 	uaddr = uvm_uarea_alloc();
547 	if (uaddr == 0) {
548 		nthreads--;
549 		return ENOMEM;
550 	}
551 
552 	/*
553 	 * From now on, we're committed to the fork and cannot fail.
554 	 */
555 	p = thread_new(curp, uaddr);
556 	atomic_setbits_int(&p->p_flag, P_THREAD);
557 	sigstkinit(&p->p_sigstk);
558 	memset(p->p_name, 0, sizeof p->p_name);
559 
560 	/* other links */
561 	p->p_p = pr;
562 
563 	/* local copies */
564 	p->p_fd		= pr->ps_fd;
565 	p->p_vmspace	= pr->ps_vmspace;
566 
567 	/*
568 	 * Finish creating the child thread.  cpu_fork() will copy
569 	 * and update the pcb and make the child ready to run.  The
570 	 * child will exit directly to user mode via child_return()
571 	 * on its first time slice and will not return here.
572 	 */
573 	cpu_fork(curp, p, stack, tcb, child_return, p);
574 
575 	p->p_tid = alloctid();
576 
577 	LIST_INSERT_HEAD(&allproc, p, p_list);
578 	LIST_INSERT_HEAD(TIDHASH(p->p_tid), p, p_hash);
579 
580 	mtx_enter(&pr->ps_mtx);
581 	TAILQ_INSERT_TAIL(&pr->ps_threads, p, p_thr_link);
582 	pr->ps_threadcnt++;
583 
584 	/*
585 	 * if somebody else wants to take us to single threaded mode,
586 	 * count ourselves in.
587 	 */
588 	if (pr->ps_single) {
589 		pr->ps_singlecnt++;
590 		atomic_setbits_int(&p->p_flag, P_SUSPSINGLE);
591 	}
592 	mtx_leave(&pr->ps_mtx);
593 
594 	/*
595 	 * Return tid to parent thread and copy it out to userspace
596 	 */
597 	*retval = tid = p->p_tid + THREAD_PID_OFFSET;
598 	if (tidptr != NULL) {
599 		if (copyout(&tid, tidptr, sizeof(tid)))
600 			psignal(curp, SIGSEGV);
601 	}
602 
603 	fork_thread_start(p, curp, 0);
604 
605 	/*
606 	 * Update stats now that we know the fork was successful.
607 	 */
608 	forkstat.cnttfork++;
609 	uvmexp.forks++;
610 	uvmexp.forks_sharevm++;
611 
612 	return 0;
613 }
614 
615 
616 /* Find an unused tid */
617 pid_t
618 alloctid(void)
619 {
620 	pid_t tid;
621 
622 	do {
623 		/* (0 .. TID_MASK+1] */
624 		tid = 1 + (arc4random() & TID_MASK);
625 	} while (tfind(tid) != NULL);
626 
627 	return (tid);
628 }
629 
630 /*
631  * Checks for current use of a pid, either as a pid or pgid.
632  */
633 pid_t oldpids[128];
634 int
635 ispidtaken(pid_t pid)
636 {
637 	uint32_t i;
638 
639 	for (i = 0; i < nitems(oldpids); i++)
640 		if (pid == oldpids[i])
641 			return (1);
642 
643 	if (prfind(pid) != NULL)
644 		return (1);
645 	if (pgfind(pid) != NULL)
646 		return (1);
647 	if (zombiefind(pid) != NULL)
648 		return (1);
649 	return (0);
650 }
651 
652 /* Find an unused pid */
653 pid_t
654 allocpid(void)
655 {
656 	static int first = 1;
657 	pid_t pid;
658 
659 	/* The first PID allocated is always 1. */
660 	if (first) {
661 		first = 0;
662 		return 1;
663 	}
664 
665 	/*
666 	 * All subsequent PIDs are chosen randomly.  We need to
667 	 * find an unused PID in the range [2, PID_MAX].
668 	 */
669 	do {
670 		pid = 2 + arc4random_uniform(PID_MAX - 1);
671 	} while (ispidtaken(pid));
672 	return pid;
673 }
674 
675 void
676 freepid(pid_t pid)
677 {
678 	static uint32_t idx;
679 
680 	oldpids[idx++ % nitems(oldpids)] = pid;
681 }
682 
683 /* Do machine independent parts of switching to a new process */
684 void
685 proc_trampoline_mi(void)
686 {
687 	struct schedstate_percpu *spc = &curcpu()->ci_schedstate;
688 	struct proc *p = curproc;
689 
690 	SCHED_ASSERT_LOCKED();
691 	clear_resched(curcpu());
692 	mtx_leave(&sched_lock);
693 	spl0();
694 
695 	SCHED_ASSERT_UNLOCKED();
696 	KERNEL_ASSERT_UNLOCKED();
697 	assertwaitok();
698 	smr_idle();
699 
700 	/* Start any optional clock interrupts needed by the thread. */
701 	if (ISSET(p->p_p->ps_flags, PS_ITIMER)) {
702 		atomic_setbits_int(&spc->spc_schedflags, SPCF_ITIMER);
703 		clockintr_advance(&spc->spc_itimer, hardclock_period);
704 	}
705 	if (ISSET(p->p_p->ps_flags, PS_PROFIL)) {
706 		atomic_setbits_int(&spc->spc_schedflags, SPCF_PROFCLOCK);
707 		clockintr_advance(&spc->spc_profclock, profclock_period);
708 	}
709 
710 	nanouptime(&spc->spc_runtime);
711 	KERNEL_LOCK();
712 }
713