xref: /freebsd/sys/kern/kern_thread.c (revision d9a42747950146bf03cda7f6e25d219253f8a57a)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
3  *
4  * Copyright (C) 2001 Julian Elischer <julian@freebsd.org>.
5  *  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice(s), this list of conditions and the following disclaimer as
12  *    the first lines of this file unmodified other than the possible
13  *    addition of one or more copyright notices.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice(s), this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  *
18  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDER(S) ``AS IS'' AND ANY
19  * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
20  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
21  * DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY
22  * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
23  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
24  * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
25  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
26  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
27  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH
28  * DAMAGE.
29  */
30 
31 #include "opt_witness.h"
32 #include "opt_hwpmc_hooks.h"
33 
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
36 
37 #include <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/kernel.h>
40 #include <sys/lock.h>
41 #include <sys/msan.h>
42 #include <sys/mutex.h>
43 #include <sys/proc.h>
44 #include <sys/bitstring.h>
45 #include <sys/epoch.h>
46 #include <sys/rangelock.h>
47 #include <sys/resourcevar.h>
48 #include <sys/sdt.h>
49 #include <sys/smp.h>
50 #include <sys/sched.h>
51 #include <sys/sleepqueue.h>
52 #include <sys/selinfo.h>
53 #include <sys/syscallsubr.h>
54 #include <sys/dtrace_bsd.h>
55 #include <sys/sysent.h>
56 #include <sys/turnstile.h>
57 #include <sys/taskqueue.h>
58 #include <sys/ktr.h>
59 #include <sys/rwlock.h>
60 #include <sys/umtxvar.h>
61 #include <sys/vmmeter.h>
62 #include <sys/cpuset.h>
63 #ifdef	HWPMC_HOOKS
64 #include <sys/pmckern.h>
65 #endif
66 #include <sys/priv.h>
67 
68 #include <security/audit/audit.h>
69 
70 #include <vm/pmap.h>
71 #include <vm/vm.h>
72 #include <vm/vm_extern.h>
73 #include <vm/uma.h>
74 #include <vm/vm_phys.h>
75 #include <sys/eventhandler.h>
76 
77 /*
78  * Asserts below verify the stability of struct thread and struct proc
79  * layout, as exposed by KBI to modules.  On head, the KBI is allowed
80  * to drift, change to the structures must be accompanied by the
81  * assert update.
82  *
83  * On the stable branches after KBI freeze, conditions must not be
84  * violated.  Typically new fields are moved to the end of the
85  * structures.
86  */
87 #ifdef __amd64__
88 _Static_assert(offsetof(struct thread, td_flags) == 0x108,
89     "struct thread KBI td_flags");
90 _Static_assert(offsetof(struct thread, td_pflags) == 0x114,
91     "struct thread KBI td_pflags");
92 _Static_assert(offsetof(struct thread, td_frame) == 0x4b0,
93     "struct thread KBI td_frame");
94 _Static_assert(offsetof(struct thread, td_emuldata) == 0x6c0,
95     "struct thread KBI td_emuldata");
96 _Static_assert(offsetof(struct proc, p_flag) == 0xb8,
97     "struct proc KBI p_flag");
98 _Static_assert(offsetof(struct proc, p_pid) == 0xc4,
99     "struct proc KBI p_pid");
100 _Static_assert(offsetof(struct proc, p_filemon) == 0x3c8,
101     "struct proc KBI p_filemon");
102 _Static_assert(offsetof(struct proc, p_comm) == 0x3e0,
103     "struct proc KBI p_comm");
104 _Static_assert(offsetof(struct proc, p_emuldata) == 0x4d0,
105     "struct proc KBI p_emuldata");
106 #endif
107 #ifdef __i386__
108 _Static_assert(offsetof(struct thread, td_flags) == 0x9c,
109     "struct thread KBI td_flags");
110 _Static_assert(offsetof(struct thread, td_pflags) == 0xa8,
111     "struct thread KBI td_pflags");
112 _Static_assert(offsetof(struct thread, td_frame) == 0x30c,
113     "struct thread KBI td_frame");
114 _Static_assert(offsetof(struct thread, td_emuldata) == 0x350,
115     "struct thread KBI td_emuldata");
116 _Static_assert(offsetof(struct proc, p_flag) == 0x6c,
117     "struct proc KBI p_flag");
118 _Static_assert(offsetof(struct proc, p_pid) == 0x78,
119     "struct proc KBI p_pid");
120 _Static_assert(offsetof(struct proc, p_filemon) == 0x270,
121     "struct proc KBI p_filemon");
122 _Static_assert(offsetof(struct proc, p_comm) == 0x284,
123     "struct proc KBI p_comm");
124 _Static_assert(offsetof(struct proc, p_emuldata) == 0x318,
125     "struct proc KBI p_emuldata");
126 #endif
127 
128 SDT_PROVIDER_DECLARE(proc);
129 SDT_PROBE_DEFINE(proc, , , lwp__exit);
130 
131 /*
132  * thread related storage.
133  */
134 static uma_zone_t thread_zone;
135 
136 struct thread_domain_data {
137 	struct thread	*tdd_zombies;
138 	int		tdd_reapticks;
139 } __aligned(CACHE_LINE_SIZE);
140 
141 static struct thread_domain_data thread_domain_data[MAXMEMDOM];
142 
143 static struct task	thread_reap_task;
144 static struct callout  	thread_reap_callout;
145 
146 static void thread_zombie(struct thread *);
147 static void thread_reap(void);
148 static void thread_reap_all(void);
149 static void thread_reap_task_cb(void *, int);
150 static void thread_reap_callout_cb(void *);
151 static int thread_unsuspend_one(struct thread *td, struct proc *p,
152     bool boundary);
153 static void thread_free_batched(struct thread *td);
154 
155 static __exclusive_cache_line struct mtx tid_lock;
156 static bitstr_t *tid_bitmap;
157 
158 static MALLOC_DEFINE(M_TIDHASH, "tidhash", "thread hash");
159 
160 static int maxthread;
161 SYSCTL_INT(_kern, OID_AUTO, maxthread, CTLFLAG_RDTUN,
162     &maxthread, 0, "Maximum number of threads");
163 
164 static __exclusive_cache_line int nthreads;
165 
166 static LIST_HEAD(tidhashhead, thread) *tidhashtbl;
167 static u_long	tidhash;
168 static u_long	tidhashlock;
169 static struct	rwlock *tidhashtbl_lock;
170 #define	TIDHASH(tid)		(&tidhashtbl[(tid) & tidhash])
171 #define	TIDHASHLOCK(tid)	(&tidhashtbl_lock[(tid) & tidhashlock])
172 
173 EVENTHANDLER_LIST_DEFINE(thread_ctor);
174 EVENTHANDLER_LIST_DEFINE(thread_dtor);
175 EVENTHANDLER_LIST_DEFINE(thread_init);
176 EVENTHANDLER_LIST_DEFINE(thread_fini);
177 
178 static bool
179 thread_count_inc_try(void)
180 {
181 	int nthreads_new;
182 
183 	nthreads_new = atomic_fetchadd_int(&nthreads, 1) + 1;
184 	if (nthreads_new >= maxthread - 100) {
185 		if (priv_check_cred(curthread->td_ucred, PRIV_MAXPROC) != 0 ||
186 		    nthreads_new >= maxthread) {
187 			atomic_subtract_int(&nthreads, 1);
188 			return (false);
189 		}
190 	}
191 	return (true);
192 }
193 
194 static bool
195 thread_count_inc(void)
196 {
197 	static struct timeval lastfail;
198 	static int curfail;
199 
200 	thread_reap();
201 	if (thread_count_inc_try()) {
202 		return (true);
203 	}
204 
205 	thread_reap_all();
206 	if (thread_count_inc_try()) {
207 		return (true);
208 	}
209 
210 	if (ppsratecheck(&lastfail, &curfail, 1)) {
211 		printf("maxthread limit exceeded by uid %u "
212 		    "(pid %d); consider increasing kern.maxthread\n",
213 		    curthread->td_ucred->cr_ruid, curproc->p_pid);
214 	}
215 	return (false);
216 }
217 
218 static void
219 thread_count_sub(int n)
220 {
221 
222 	atomic_subtract_int(&nthreads, n);
223 }
224 
225 static void
226 thread_count_dec(void)
227 {
228 
229 	thread_count_sub(1);
230 }
231 
232 static lwpid_t
233 tid_alloc(void)
234 {
235 	static lwpid_t trytid;
236 	lwpid_t tid;
237 
238 	mtx_lock(&tid_lock);
239 	/*
240 	 * It is an invariant that the bitmap is big enough to hold maxthread
241 	 * IDs. If we got to this point there has to be at least one free.
242 	 */
243 	if (trytid >= maxthread)
244 		trytid = 0;
245 	bit_ffc_at(tid_bitmap, trytid, maxthread, &tid);
246 	if (tid == -1) {
247 		KASSERT(trytid != 0, ("unexpectedly ran out of IDs"));
248 		trytid = 0;
249 		bit_ffc_at(tid_bitmap, trytid, maxthread, &tid);
250 		KASSERT(tid != -1, ("unexpectedly ran out of IDs"));
251 	}
252 	bit_set(tid_bitmap, tid);
253 	trytid = tid + 1;
254 	mtx_unlock(&tid_lock);
255 	return (tid + NO_PID);
256 }
257 
258 static void
259 tid_free_locked(lwpid_t rtid)
260 {
261 	lwpid_t tid;
262 
263 	mtx_assert(&tid_lock, MA_OWNED);
264 	KASSERT(rtid >= NO_PID,
265 	    ("%s: invalid tid %d\n", __func__, rtid));
266 	tid = rtid - NO_PID;
267 	KASSERT(bit_test(tid_bitmap, tid) != 0,
268 	    ("thread ID %d not allocated\n", rtid));
269 	bit_clear(tid_bitmap, tid);
270 }
271 
272 static void
273 tid_free(lwpid_t rtid)
274 {
275 
276 	mtx_lock(&tid_lock);
277 	tid_free_locked(rtid);
278 	mtx_unlock(&tid_lock);
279 }
280 
281 static void
282 tid_free_batch(lwpid_t *batch, int n)
283 {
284 	int i;
285 
286 	mtx_lock(&tid_lock);
287 	for (i = 0; i < n; i++) {
288 		tid_free_locked(batch[i]);
289 	}
290 	mtx_unlock(&tid_lock);
291 }
292 
293 /*
294  * Batching for thread reapping.
295  */
296 struct tidbatch {
297 	lwpid_t tab[16];
298 	int n;
299 };
300 
301 static void
302 tidbatch_prep(struct tidbatch *tb)
303 {
304 
305 	tb->n = 0;
306 }
307 
308 static void
309 tidbatch_add(struct tidbatch *tb, struct thread *td)
310 {
311 
312 	KASSERT(tb->n < nitems(tb->tab),
313 	    ("%s: count too high %d", __func__, tb->n));
314 	tb->tab[tb->n] = td->td_tid;
315 	tb->n++;
316 }
317 
318 static void
319 tidbatch_process(struct tidbatch *tb)
320 {
321 
322 	KASSERT(tb->n <= nitems(tb->tab),
323 	    ("%s: count too high %d", __func__, tb->n));
324 	if (tb->n == nitems(tb->tab)) {
325 		tid_free_batch(tb->tab, tb->n);
326 		tb->n = 0;
327 	}
328 }
329 
330 static void
331 tidbatch_final(struct tidbatch *tb)
332 {
333 
334 	KASSERT(tb->n <= nitems(tb->tab),
335 	    ("%s: count too high %d", __func__, tb->n));
336 	if (tb->n != 0) {
337 		tid_free_batch(tb->tab, tb->n);
338 	}
339 }
340 
341 /*
342  * Prepare a thread for use.
343  */
344 static int
345 thread_ctor(void *mem, int size, void *arg, int flags)
346 {
347 	struct thread	*td;
348 
349 	td = (struct thread *)mem;
350 	TD_SET_STATE(td, TDS_INACTIVE);
351 	td->td_lastcpu = td->td_oncpu = NOCPU;
352 
353 	/*
354 	 * Note that td_critnest begins life as 1 because the thread is not
355 	 * running and is thereby implicitly waiting to be on the receiving
356 	 * end of a context switch.
357 	 */
358 	td->td_critnest = 1;
359 	td->td_lend_user_pri = PRI_MAX;
360 #ifdef AUDIT
361 	audit_thread_alloc(td);
362 #endif
363 #ifdef KDTRACE_HOOKS
364 	kdtrace_thread_ctor(td);
365 #endif
366 	umtx_thread_alloc(td);
367 	MPASS(td->td_sel == NULL);
368 	return (0);
369 }
370 
371 /*
372  * Reclaim a thread after use.
373  */
374 static void
375 thread_dtor(void *mem, int size, void *arg)
376 {
377 	struct thread *td;
378 
379 	td = (struct thread *)mem;
380 
381 #ifdef INVARIANTS
382 	/* Verify that this thread is in a safe state to free. */
383 	switch (TD_GET_STATE(td)) {
384 	case TDS_INHIBITED:
385 	case TDS_RUNNING:
386 	case TDS_CAN_RUN:
387 	case TDS_RUNQ:
388 		/*
389 		 * We must never unlink a thread that is in one of
390 		 * these states, because it is currently active.
391 		 */
392 		panic("bad state for thread unlinking");
393 		/* NOTREACHED */
394 	case TDS_INACTIVE:
395 		break;
396 	default:
397 		panic("bad thread state");
398 		/* NOTREACHED */
399 	}
400 #endif
401 #ifdef AUDIT
402 	audit_thread_free(td);
403 #endif
404 #ifdef KDTRACE_HOOKS
405 	kdtrace_thread_dtor(td);
406 #endif
407 	/* Free all OSD associated to this thread. */
408 	osd_thread_exit(td);
409 	ast_kclear(td);
410 	seltdfini(td);
411 }
412 
413 /*
414  * Initialize type-stable parts of a thread (when newly created).
415  */
416 static int
417 thread_init(void *mem, int size, int flags)
418 {
419 	struct thread *td;
420 
421 	td = (struct thread *)mem;
422 
423 	td->td_allocdomain = vm_phys_domain(vtophys(td));
424 	td->td_sleepqueue = sleepq_alloc();
425 	td->td_turnstile = turnstile_alloc();
426 	td->td_rlqe = NULL;
427 	EVENTHANDLER_DIRECT_INVOKE(thread_init, td);
428 	umtx_thread_init(td);
429 	td->td_kstack = 0;
430 	td->td_sel = NULL;
431 	return (0);
432 }
433 
434 /*
435  * Tear down type-stable parts of a thread (just before being discarded).
436  */
437 static void
438 thread_fini(void *mem, int size)
439 {
440 	struct thread *td;
441 
442 	td = (struct thread *)mem;
443 	EVENTHANDLER_DIRECT_INVOKE(thread_fini, td);
444 	rlqentry_free(td->td_rlqe);
445 	turnstile_free(td->td_turnstile);
446 	sleepq_free(td->td_sleepqueue);
447 	umtx_thread_fini(td);
448 	MPASS(td->td_sel == NULL);
449 }
450 
451 /*
452  * For a newly created process,
453  * link up all the structures and its initial threads etc.
454  * called from:
455  * {arch}/{arch}/machdep.c   {arch}_init(), init386() etc.
456  * proc_dtor() (should go away)
457  * proc_init()
458  */
459 void
460 proc_linkup0(struct proc *p, struct thread *td)
461 {
462 	TAILQ_INIT(&p->p_threads);	     /* all threads in proc */
463 	proc_linkup(p, td);
464 }
465 
466 void
467 proc_linkup(struct proc *p, struct thread *td)
468 {
469 
470 	sigqueue_init(&p->p_sigqueue, p);
471 	p->p_ksi = ksiginfo_alloc(M_WAITOK);
472 	if (p->p_ksi != NULL) {
473 		/* XXX p_ksi may be null if ksiginfo zone is not ready */
474 		p->p_ksi->ksi_flags = KSI_EXT | KSI_INS;
475 	}
476 	LIST_INIT(&p->p_mqnotifier);
477 	p->p_numthreads = 0;
478 	thread_link(td, p);
479 }
480 
481 static void
482 ast_suspend(struct thread *td, int tda __unused)
483 {
484 	struct proc *p;
485 
486 	p = td->td_proc;
487 	/*
488 	 * We need to check to see if we have to exit or wait due to a
489 	 * single threading requirement or some other STOP condition.
490 	 */
491 	PROC_LOCK(p);
492 	thread_suspend_check(0);
493 	PROC_UNLOCK(p);
494 }
495 
496 extern int max_threads_per_proc;
497 
498 /*
499  * Initialize global thread allocation resources.
500  */
501 void
502 threadinit(void)
503 {
504 	u_long i;
505 	lwpid_t tid0;
506 
507 	/*
508 	 * Place an upper limit on threads which can be allocated.
509 	 *
510 	 * Note that other factors may make the de facto limit much lower.
511 	 *
512 	 * Platform limits are somewhat arbitrary but deemed "more than good
513 	 * enough" for the foreseable future.
514 	 */
515 	if (maxthread == 0) {
516 #ifdef _LP64
517 		maxthread = MIN(maxproc * max_threads_per_proc, 1000000);
518 #else
519 		maxthread = MIN(maxproc * max_threads_per_proc, 100000);
520 #endif
521 	}
522 
523 	mtx_init(&tid_lock, "TID lock", NULL, MTX_DEF);
524 	tid_bitmap = bit_alloc(maxthread, M_TIDHASH, M_WAITOK);
525 	/*
526 	 * Handle thread0.
527 	 */
528 	thread_count_inc();
529 	tid0 = tid_alloc();
530 	if (tid0 != THREAD0_TID)
531 		panic("tid0 %d != %d\n", tid0, THREAD0_TID);
532 
533 	thread_zone = uma_zcreate("THREAD", sched_sizeof_thread(),
534 	    thread_ctor, thread_dtor, thread_init, thread_fini,
535 	    32 - 1, UMA_ZONE_NOFREE);
536 	tidhashtbl = hashinit(maxproc / 2, M_TIDHASH, &tidhash);
537 	tidhashlock = (tidhash + 1) / 64;
538 	if (tidhashlock > 0)
539 		tidhashlock--;
540 	tidhashtbl_lock = malloc(sizeof(*tidhashtbl_lock) * (tidhashlock + 1),
541 	    M_TIDHASH, M_WAITOK | M_ZERO);
542 	for (i = 0; i < tidhashlock + 1; i++)
543 		rw_init(&tidhashtbl_lock[i], "tidhash");
544 
545 	TASK_INIT(&thread_reap_task, 0, thread_reap_task_cb, NULL);
546 	callout_init(&thread_reap_callout, 1);
547 	callout_reset(&thread_reap_callout, 5 * hz,
548 	    thread_reap_callout_cb, NULL);
549 	ast_register(TDA_SUSPEND, ASTR_ASTF_REQUIRED, 0, ast_suspend);
550 }
551 
552 /*
553  * Place an unused thread on the zombie list.
554  */
555 void
556 thread_zombie(struct thread *td)
557 {
558 	struct thread_domain_data *tdd;
559 	struct thread *ztd;
560 
561 	tdd = &thread_domain_data[td->td_allocdomain];
562 	ztd = atomic_load_ptr(&tdd->tdd_zombies);
563 	for (;;) {
564 		td->td_zombie = ztd;
565 		if (atomic_fcmpset_rel_ptr((uintptr_t *)&tdd->tdd_zombies,
566 		    (uintptr_t *)&ztd, (uintptr_t)td))
567 			break;
568 		continue;
569 	}
570 }
571 
572 /*
573  * Release a thread that has exited after cpu_throw().
574  */
575 void
576 thread_stash(struct thread *td)
577 {
578 	atomic_subtract_rel_int(&td->td_proc->p_exitthreads, 1);
579 	thread_zombie(td);
580 }
581 
582 /*
583  * Reap zombies from passed domain.
584  */
585 static void
586 thread_reap_domain(struct thread_domain_data *tdd)
587 {
588 	struct thread *itd, *ntd;
589 	struct tidbatch tidbatch;
590 	struct credbatch credbatch;
591 	int tdcount;
592 	struct plimit *lim;
593 	int limcount;
594 
595 	/*
596 	 * Reading upfront is pessimal if followed by concurrent atomic_swap,
597 	 * but most of the time the list is empty.
598 	 */
599 	if (tdd->tdd_zombies == NULL)
600 		return;
601 
602 	itd = (struct thread *)atomic_swap_ptr((uintptr_t *)&tdd->tdd_zombies,
603 	    (uintptr_t)NULL);
604 	if (itd == NULL)
605 		return;
606 
607 	/*
608 	 * Multiple CPUs can get here, the race is fine as ticks is only
609 	 * advisory.
610 	 */
611 	tdd->tdd_reapticks = ticks;
612 
613 	tidbatch_prep(&tidbatch);
614 	credbatch_prep(&credbatch);
615 	tdcount = 0;
616 	lim = NULL;
617 	limcount = 0;
618 
619 	while (itd != NULL) {
620 		ntd = itd->td_zombie;
621 		EVENTHANDLER_DIRECT_INVOKE(thread_dtor, itd);
622 		tidbatch_add(&tidbatch, itd);
623 		credbatch_add(&credbatch, itd);
624 		MPASS(itd->td_limit != NULL);
625 		if (lim != itd->td_limit) {
626 			if (limcount != 0) {
627 				lim_freen(lim, limcount);
628 				limcount = 0;
629 			}
630 		}
631 		lim = itd->td_limit;
632 		limcount++;
633 		thread_free_batched(itd);
634 		tidbatch_process(&tidbatch);
635 		credbatch_process(&credbatch);
636 		tdcount++;
637 		if (tdcount == 32) {
638 			thread_count_sub(tdcount);
639 			tdcount = 0;
640 		}
641 		itd = ntd;
642 	}
643 
644 	tidbatch_final(&tidbatch);
645 	credbatch_final(&credbatch);
646 	if (tdcount != 0) {
647 		thread_count_sub(tdcount);
648 	}
649 	MPASS(limcount != 0);
650 	lim_freen(lim, limcount);
651 }
652 
653 /*
654  * Reap zombies from all domains.
655  */
656 static void
657 thread_reap_all(void)
658 {
659 	struct thread_domain_data *tdd;
660 	int i, domain;
661 
662 	domain = PCPU_GET(domain);
663 	for (i = 0; i < vm_ndomains; i++) {
664 		tdd = &thread_domain_data[(i + domain) % vm_ndomains];
665 		thread_reap_domain(tdd);
666 	}
667 }
668 
669 /*
670  * Reap zombies from local domain.
671  */
672 static void
673 thread_reap(void)
674 {
675 	struct thread_domain_data *tdd;
676 	int domain;
677 
678 	domain = PCPU_GET(domain);
679 	tdd = &thread_domain_data[domain];
680 
681 	thread_reap_domain(tdd);
682 }
683 
684 static void
685 thread_reap_task_cb(void *arg __unused, int pending __unused)
686 {
687 
688 	thread_reap_all();
689 }
690 
691 static void
692 thread_reap_callout_cb(void *arg __unused)
693 {
694 	struct thread_domain_data *tdd;
695 	int i, cticks, lticks;
696 	bool wantreap;
697 
698 	wantreap = false;
699 	cticks = atomic_load_int(&ticks);
700 	for (i = 0; i < vm_ndomains; i++) {
701 		tdd = &thread_domain_data[i];
702 		lticks = tdd->tdd_reapticks;
703 		if (tdd->tdd_zombies != NULL &&
704 		    (u_int)(cticks - lticks) > 5 * hz) {
705 			wantreap = true;
706 			break;
707 		}
708 	}
709 
710 	if (wantreap)
711 		taskqueue_enqueue(taskqueue_thread, &thread_reap_task);
712 	callout_reset(&thread_reap_callout, 5 * hz,
713 	    thread_reap_callout_cb, NULL);
714 }
715 
716 /*
717  * Calling this function guarantees that any thread that exited before
718  * the call is reaped when the function returns.  By 'exited' we mean
719  * a thread removed from the process linkage with thread_unlink().
720  * Practically this means that caller must lock/unlock corresponding
721  * process lock before the call, to synchronize with thread_exit().
722  */
723 void
724 thread_reap_barrier(void)
725 {
726 	struct task *t;
727 
728 	/*
729 	 * First do context switches to each CPU to ensure that all
730 	 * PCPU pc_deadthreads are moved to zombie list.
731 	 */
732 	quiesce_all_cpus("", PDROP);
733 
734 	/*
735 	 * Second, fire the task in the same thread as normal
736 	 * thread_reap() is done, to serialize reaping.
737 	 */
738 	t = malloc(sizeof(*t), M_TEMP, M_WAITOK);
739 	TASK_INIT(t, 0, thread_reap_task_cb, t);
740 	taskqueue_enqueue(taskqueue_thread, t);
741 	taskqueue_drain(taskqueue_thread, t);
742 	free(t, M_TEMP);
743 }
744 
745 /*
746  * Allocate a thread.
747  */
748 struct thread *
749 thread_alloc(int pages)
750 {
751 	struct thread *td;
752 	lwpid_t tid;
753 
754 	if (!thread_count_inc()) {
755 		return (NULL);
756 	}
757 
758 	tid = tid_alloc();
759 	td = uma_zalloc(thread_zone, M_WAITOK);
760 	KASSERT(td->td_kstack == 0, ("thread_alloc got thread with kstack"));
761 	if (!vm_thread_new(td, pages)) {
762 		uma_zfree(thread_zone, td);
763 		tid_free(tid);
764 		thread_count_dec();
765 		return (NULL);
766 	}
767 	td->td_tid = tid;
768 	bzero(&td->td_sa.args, sizeof(td->td_sa.args));
769 	kmsan_thread_alloc(td);
770 	cpu_thread_alloc(td);
771 	EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
772 	return (td);
773 }
774 
775 int
776 thread_alloc_stack(struct thread *td, int pages)
777 {
778 
779 	KASSERT(td->td_kstack == 0,
780 	    ("thread_alloc_stack called on a thread with kstack"));
781 	if (!vm_thread_new(td, pages))
782 		return (0);
783 	cpu_thread_alloc(td);
784 	return (1);
785 }
786 
787 /*
788  * Deallocate a thread.
789  */
790 static void
791 thread_free_batched(struct thread *td)
792 {
793 
794 	lock_profile_thread_exit(td);
795 	if (td->td_cpuset)
796 		cpuset_rel(td->td_cpuset);
797 	td->td_cpuset = NULL;
798 	cpu_thread_free(td);
799 	if (td->td_kstack != 0)
800 		vm_thread_dispose(td);
801 	callout_drain(&td->td_slpcallout);
802 	/*
803 	 * Freeing handled by the caller.
804 	 */
805 	td->td_tid = -1;
806 	kmsan_thread_free(td);
807 	uma_zfree(thread_zone, td);
808 }
809 
810 void
811 thread_free(struct thread *td)
812 {
813 	lwpid_t tid;
814 
815 	EVENTHANDLER_DIRECT_INVOKE(thread_dtor, td);
816 	tid = td->td_tid;
817 	thread_free_batched(td);
818 	tid_free(tid);
819 	thread_count_dec();
820 }
821 
822 void
823 thread_cow_get_proc(struct thread *newtd, struct proc *p)
824 {
825 
826 	PROC_LOCK_ASSERT(p, MA_OWNED);
827 	newtd->td_realucred = crcowget(p->p_ucred);
828 	newtd->td_ucred = newtd->td_realucred;
829 	newtd->td_limit = lim_hold(p->p_limit);
830 	newtd->td_cowgen = p->p_cowgen;
831 }
832 
833 void
834 thread_cow_get(struct thread *newtd, struct thread *td)
835 {
836 
837 	MPASS(td->td_realucred == td->td_ucred);
838 	newtd->td_realucred = crcowget(td->td_realucred);
839 	newtd->td_ucred = newtd->td_realucred;
840 	newtd->td_limit = lim_hold(td->td_limit);
841 	newtd->td_cowgen = td->td_cowgen;
842 }
843 
844 void
845 thread_cow_free(struct thread *td)
846 {
847 
848 	if (td->td_realucred != NULL)
849 		crcowfree(td);
850 	if (td->td_limit != NULL)
851 		lim_free(td->td_limit);
852 }
853 
854 void
855 thread_cow_update(struct thread *td)
856 {
857 	struct proc *p;
858 	struct ucred *oldcred;
859 	struct plimit *oldlimit;
860 
861 	p = td->td_proc;
862 	PROC_LOCK(p);
863 	oldcred = crcowsync();
864 	oldlimit = lim_cowsync();
865 	td->td_cowgen = p->p_cowgen;
866 	PROC_UNLOCK(p);
867 	if (oldcred != NULL)
868 		crfree(oldcred);
869 	if (oldlimit != NULL)
870 		lim_free(oldlimit);
871 }
872 
873 void
874 thread_cow_synced(struct thread *td)
875 {
876 	struct proc *p;
877 
878 	p = td->td_proc;
879 	PROC_LOCK_ASSERT(p, MA_OWNED);
880 	MPASS(td->td_cowgen != p->p_cowgen);
881 	MPASS(td->td_ucred == p->p_ucred);
882 	MPASS(td->td_limit == p->p_limit);
883 	td->td_cowgen = p->p_cowgen;
884 }
885 
886 /*
887  * Discard the current thread and exit from its context.
888  * Always called with scheduler locked.
889  *
890  * Because we can't free a thread while we're operating under its context,
891  * push the current thread into our CPU's deadthread holder. This means
892  * we needn't worry about someone else grabbing our context before we
893  * do a cpu_throw().
894  */
895 void
896 thread_exit(void)
897 {
898 	uint64_t runtime, new_switchtime;
899 	struct thread *td;
900 	struct thread *td2;
901 	struct proc *p;
902 	int wakeup_swapper;
903 
904 	td = curthread;
905 	p = td->td_proc;
906 
907 	PROC_SLOCK_ASSERT(p, MA_OWNED);
908 	mtx_assert(&Giant, MA_NOTOWNED);
909 
910 	PROC_LOCK_ASSERT(p, MA_OWNED);
911 	KASSERT(p != NULL, ("thread exiting without a process"));
912 	CTR3(KTR_PROC, "thread_exit: thread %p (pid %ld, %s)", td,
913 	    (long)p->p_pid, td->td_name);
914 	SDT_PROBE0(proc, , , lwp__exit);
915 	KASSERT(TAILQ_EMPTY(&td->td_sigqueue.sq_list), ("signal pending"));
916 	MPASS(td->td_realucred == td->td_ucred);
917 
918 	/*
919 	 * drop FPU & debug register state storage, or any other
920 	 * architecture specific resources that
921 	 * would not be on a new untouched process.
922 	 */
923 	cpu_thread_exit(td);
924 
925 	/*
926 	 * The last thread is left attached to the process
927 	 * So that the whole bundle gets recycled. Skip
928 	 * all this stuff if we never had threads.
929 	 * EXIT clears all sign of other threads when
930 	 * it goes to single threading, so the last thread always
931 	 * takes the short path.
932 	 */
933 	if (p->p_flag & P_HADTHREADS) {
934 		if (p->p_numthreads > 1) {
935 			atomic_add_int(&td->td_proc->p_exitthreads, 1);
936 			thread_unlink(td);
937 			td2 = FIRST_THREAD_IN_PROC(p);
938 			sched_exit_thread(td2, td);
939 
940 			/*
941 			 * The test below is NOT true if we are the
942 			 * sole exiting thread. P_STOPPED_SINGLE is unset
943 			 * in exit1() after it is the only survivor.
944 			 */
945 			if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
946 				if (p->p_numthreads == p->p_suspcount) {
947 					thread_lock(p->p_singlethread);
948 					wakeup_swapper = thread_unsuspend_one(
949 						p->p_singlethread, p, false);
950 					if (wakeup_swapper)
951 						kick_proc0();
952 				}
953 			}
954 
955 			PCPU_SET(deadthread, td);
956 		} else {
957 			/*
958 			 * The last thread is exiting.. but not through exit()
959 			 */
960 			panic ("thread_exit: Last thread exiting on its own");
961 		}
962 	}
963 #ifdef	HWPMC_HOOKS
964 	/*
965 	 * If this thread is part of a process that is being tracked by hwpmc(4),
966 	 * inform the module of the thread's impending exit.
967 	 */
968 	if (PMC_PROC_IS_USING_PMCS(td->td_proc)) {
969 		PMC_SWITCH_CONTEXT(td, PMC_FN_CSW_OUT);
970 		PMC_CALL_HOOK_UNLOCKED(td, PMC_FN_THR_EXIT, NULL);
971 	} else if (PMC_SYSTEM_SAMPLING_ACTIVE())
972 		PMC_CALL_HOOK_UNLOCKED(td, PMC_FN_THR_EXIT_LOG, NULL);
973 #endif
974 	PROC_UNLOCK(p);
975 	PROC_STATLOCK(p);
976 	thread_lock(td);
977 	PROC_SUNLOCK(p);
978 
979 	/* Do the same timestamp bookkeeping that mi_switch() would do. */
980 	new_switchtime = cpu_ticks();
981 	runtime = new_switchtime - PCPU_GET(switchtime);
982 	td->td_runtime += runtime;
983 	td->td_incruntime += runtime;
984 	PCPU_SET(switchtime, new_switchtime);
985 	PCPU_SET(switchticks, ticks);
986 	VM_CNT_INC(v_swtch);
987 
988 	/* Save our resource usage in our process. */
989 	td->td_ru.ru_nvcsw++;
990 	ruxagg_locked(p, td);
991 	rucollect(&p->p_ru, &td->td_ru);
992 	PROC_STATUNLOCK(p);
993 
994 	TD_SET_STATE(td, TDS_INACTIVE);
995 #ifdef WITNESS
996 	witness_thread_exit(td);
997 #endif
998 	CTR1(KTR_PROC, "thread_exit: cpu_throw() thread %p", td);
999 	sched_throw(td);
1000 	panic("I'm a teapot!");
1001 	/* NOTREACHED */
1002 }
1003 
1004 /*
1005  * Do any thread specific cleanups that may be needed in wait()
1006  * called with Giant, proc and schedlock not held.
1007  */
1008 void
1009 thread_wait(struct proc *p)
1010 {
1011 	struct thread *td;
1012 
1013 	mtx_assert(&Giant, MA_NOTOWNED);
1014 	KASSERT(p->p_numthreads == 1, ("multiple threads in thread_wait()"));
1015 	KASSERT(p->p_exitthreads == 0, ("p_exitthreads leaking"));
1016 	td = FIRST_THREAD_IN_PROC(p);
1017 	/* Lock the last thread so we spin until it exits cpu_throw(). */
1018 	thread_lock(td);
1019 	thread_unlock(td);
1020 	lock_profile_thread_exit(td);
1021 	cpuset_rel(td->td_cpuset);
1022 	td->td_cpuset = NULL;
1023 	cpu_thread_clean(td);
1024 	thread_cow_free(td);
1025 	callout_drain(&td->td_slpcallout);
1026 	thread_reap();	/* check for zombie threads etc. */
1027 }
1028 
1029 /*
1030  * Link a thread to a process.
1031  * set up anything that needs to be initialized for it to
1032  * be used by the process.
1033  */
1034 void
1035 thread_link(struct thread *td, struct proc *p)
1036 {
1037 
1038 	/*
1039 	 * XXX This can't be enabled because it's called for proc0 before
1040 	 * its lock has been created.
1041 	 * PROC_LOCK_ASSERT(p, MA_OWNED);
1042 	 */
1043 	TD_SET_STATE(td, TDS_INACTIVE);
1044 	td->td_proc     = p;
1045 	td->td_flags    = TDF_INMEM;
1046 
1047 	LIST_INIT(&td->td_contested);
1048 	LIST_INIT(&td->td_lprof[0]);
1049 	LIST_INIT(&td->td_lprof[1]);
1050 #ifdef EPOCH_TRACE
1051 	SLIST_INIT(&td->td_epochs);
1052 #endif
1053 	sigqueue_init(&td->td_sigqueue, p);
1054 	callout_init(&td->td_slpcallout, 1);
1055 	TAILQ_INSERT_TAIL(&p->p_threads, td, td_plist);
1056 	p->p_numthreads++;
1057 }
1058 
1059 /*
1060  * Called from:
1061  *  thread_exit()
1062  */
1063 void
1064 thread_unlink(struct thread *td)
1065 {
1066 	struct proc *p = td->td_proc;
1067 
1068 	PROC_LOCK_ASSERT(p, MA_OWNED);
1069 #ifdef EPOCH_TRACE
1070 	MPASS(SLIST_EMPTY(&td->td_epochs));
1071 #endif
1072 
1073 	TAILQ_REMOVE(&p->p_threads, td, td_plist);
1074 	p->p_numthreads--;
1075 	/* could clear a few other things here */
1076 	/* Must  NOT clear links to proc! */
1077 }
1078 
1079 static int
1080 calc_remaining(struct proc *p, int mode)
1081 {
1082 	int remaining;
1083 
1084 	PROC_LOCK_ASSERT(p, MA_OWNED);
1085 	PROC_SLOCK_ASSERT(p, MA_OWNED);
1086 	if (mode == SINGLE_EXIT)
1087 		remaining = p->p_numthreads;
1088 	else if (mode == SINGLE_BOUNDARY)
1089 		remaining = p->p_numthreads - p->p_boundary_count;
1090 	else if (mode == SINGLE_NO_EXIT || mode == SINGLE_ALLPROC)
1091 		remaining = p->p_numthreads - p->p_suspcount;
1092 	else
1093 		panic("calc_remaining: wrong mode %d", mode);
1094 	return (remaining);
1095 }
1096 
1097 static int
1098 remain_for_mode(int mode)
1099 {
1100 
1101 	return (mode == SINGLE_ALLPROC ? 0 : 1);
1102 }
1103 
1104 static int
1105 weed_inhib(int mode, struct thread *td2, struct proc *p)
1106 {
1107 	int wakeup_swapper;
1108 
1109 	PROC_LOCK_ASSERT(p, MA_OWNED);
1110 	PROC_SLOCK_ASSERT(p, MA_OWNED);
1111 	THREAD_LOCK_ASSERT(td2, MA_OWNED);
1112 
1113 	wakeup_swapper = 0;
1114 
1115 	/*
1116 	 * Since the thread lock is dropped by the scheduler we have
1117 	 * to retry to check for races.
1118 	 */
1119 restart:
1120 	switch (mode) {
1121 	case SINGLE_EXIT:
1122 		if (TD_IS_SUSPENDED(td2)) {
1123 			wakeup_swapper |= thread_unsuspend_one(td2, p, true);
1124 			thread_lock(td2);
1125 			goto restart;
1126 		}
1127 		if (TD_CAN_ABORT(td2)) {
1128 			wakeup_swapper |= sleepq_abort(td2, EINTR);
1129 			return (wakeup_swapper);
1130 		}
1131 		break;
1132 	case SINGLE_BOUNDARY:
1133 	case SINGLE_NO_EXIT:
1134 		if (TD_IS_SUSPENDED(td2) &&
1135 		    (td2->td_flags & TDF_BOUNDARY) == 0) {
1136 			wakeup_swapper |= thread_unsuspend_one(td2, p, false);
1137 			thread_lock(td2);
1138 			goto restart;
1139 		}
1140 		if (TD_CAN_ABORT(td2)) {
1141 			wakeup_swapper |= sleepq_abort(td2, ERESTART);
1142 			return (wakeup_swapper);
1143 		}
1144 		break;
1145 	case SINGLE_ALLPROC:
1146 		/*
1147 		 * ALLPROC suspend tries to avoid spurious EINTR for
1148 		 * threads sleeping interruptable, by suspending the
1149 		 * thread directly, similarly to sig_suspend_threads().
1150 		 * Since such sleep is not neccessary performed at the user
1151 		 * boundary, TDF_ALLPROCSUSP is used to avoid immediate
1152 		 * un-suspend.
1153 		 */
1154 		if (TD_IS_SUSPENDED(td2) &&
1155 		    (td2->td_flags & TDF_ALLPROCSUSP) == 0) {
1156 			wakeup_swapper |= thread_unsuspend_one(td2, p, false);
1157 			thread_lock(td2);
1158 			goto restart;
1159 		}
1160 		if (TD_CAN_ABORT(td2)) {
1161 			td2->td_flags |= TDF_ALLPROCSUSP;
1162 			wakeup_swapper |= sleepq_abort(td2, ERESTART);
1163 			return (wakeup_swapper);
1164 		}
1165 		break;
1166 	default:
1167 		break;
1168 	}
1169 	thread_unlock(td2);
1170 	return (wakeup_swapper);
1171 }
1172 
1173 /*
1174  * Enforce single-threading.
1175  *
1176  * Returns 1 if the caller must abort (another thread is waiting to
1177  * exit the process or similar). Process is locked!
1178  * Returns 0 when you are successfully the only thread running.
1179  * A process has successfully single threaded in the suspend mode when
1180  * There are no threads in user mode. Threads in the kernel must be
1181  * allowed to continue until they get to the user boundary. They may even
1182  * copy out their return values and data before suspending. They may however be
1183  * accelerated in reaching the user boundary as we will wake up
1184  * any sleeping threads that are interruptable. (PCATCH).
1185  */
1186 int
1187 thread_single(struct proc *p, int mode)
1188 {
1189 	struct thread *td;
1190 	struct thread *td2;
1191 	int remaining, wakeup_swapper;
1192 
1193 	td = curthread;
1194 	KASSERT(mode == SINGLE_EXIT || mode == SINGLE_BOUNDARY ||
1195 	    mode == SINGLE_ALLPROC || mode == SINGLE_NO_EXIT,
1196 	    ("invalid mode %d", mode));
1197 	/*
1198 	 * If allowing non-ALLPROC singlethreading for non-curproc
1199 	 * callers, calc_remaining() and remain_for_mode() should be
1200 	 * adjusted to also account for td->td_proc != p.  For now
1201 	 * this is not implemented because it is not used.
1202 	 */
1203 	KASSERT((mode == SINGLE_ALLPROC && td->td_proc != p) ||
1204 	    (mode != SINGLE_ALLPROC && td->td_proc == p),
1205 	    ("mode %d proc %p curproc %p", mode, p, td->td_proc));
1206 	mtx_assert(&Giant, MA_NOTOWNED);
1207 	PROC_LOCK_ASSERT(p, MA_OWNED);
1208 
1209 	/*
1210 	 * Is someone already single threading?
1211 	 * Or may be singlethreading is not needed at all.
1212 	 */
1213 	if (mode == SINGLE_ALLPROC) {
1214 		while ((p->p_flag & P_STOPPED_SINGLE) != 0) {
1215 			if ((p->p_flag2 & P2_WEXIT) != 0)
1216 				return (1);
1217 			msleep(&p->p_flag, &p->p_mtx, PCATCH, "thrsgl", 0);
1218 		}
1219 	} else if ((p->p_flag & P_HADTHREADS) == 0)
1220 		return (0);
1221 	if (p->p_singlethread != NULL && p->p_singlethread != td)
1222 		return (1);
1223 
1224 	if (mode == SINGLE_EXIT) {
1225 		p->p_flag |= P_SINGLE_EXIT;
1226 		p->p_flag &= ~P_SINGLE_BOUNDARY;
1227 	} else {
1228 		p->p_flag &= ~P_SINGLE_EXIT;
1229 		if (mode == SINGLE_BOUNDARY)
1230 			p->p_flag |= P_SINGLE_BOUNDARY;
1231 		else
1232 			p->p_flag &= ~P_SINGLE_BOUNDARY;
1233 	}
1234 	if (mode == SINGLE_ALLPROC)
1235 		p->p_flag |= P_TOTAL_STOP;
1236 	p->p_flag |= P_STOPPED_SINGLE;
1237 	PROC_SLOCK(p);
1238 	p->p_singlethread = td;
1239 	remaining = calc_remaining(p, mode);
1240 	while (remaining != remain_for_mode(mode)) {
1241 		if (P_SHOULDSTOP(p) != P_STOPPED_SINGLE)
1242 			goto stopme;
1243 		wakeup_swapper = 0;
1244 		FOREACH_THREAD_IN_PROC(p, td2) {
1245 			if (td2 == td)
1246 				continue;
1247 			thread_lock(td2);
1248 			ast_sched_locked(td2, TDA_SUSPEND);
1249 			if (TD_IS_INHIBITED(td2)) {
1250 				wakeup_swapper |= weed_inhib(mode, td2, p);
1251 #ifdef SMP
1252 			} else if (TD_IS_RUNNING(td2)) {
1253 				forward_signal(td2);
1254 				thread_unlock(td2);
1255 #endif
1256 			} else
1257 				thread_unlock(td2);
1258 		}
1259 		if (wakeup_swapper)
1260 			kick_proc0();
1261 		remaining = calc_remaining(p, mode);
1262 
1263 		/*
1264 		 * Maybe we suspended some threads.. was it enough?
1265 		 */
1266 		if (remaining == remain_for_mode(mode))
1267 			break;
1268 
1269 stopme:
1270 		/*
1271 		 * Wake us up when everyone else has suspended.
1272 		 * In the mean time we suspend as well.
1273 		 */
1274 		thread_suspend_switch(td, p);
1275 		remaining = calc_remaining(p, mode);
1276 	}
1277 	if (mode == SINGLE_EXIT) {
1278 		/*
1279 		 * Convert the process to an unthreaded process.  The
1280 		 * SINGLE_EXIT is called by exit1() or execve(), in
1281 		 * both cases other threads must be retired.
1282 		 */
1283 		KASSERT(p->p_numthreads == 1, ("Unthreading with >1 threads"));
1284 		p->p_singlethread = NULL;
1285 		p->p_flag &= ~(P_STOPPED_SINGLE | P_SINGLE_EXIT | P_HADTHREADS);
1286 
1287 		/*
1288 		 * Wait for any remaining threads to exit cpu_throw().
1289 		 */
1290 		while (p->p_exitthreads != 0) {
1291 			PROC_SUNLOCK(p);
1292 			PROC_UNLOCK(p);
1293 			sched_relinquish(td);
1294 			PROC_LOCK(p);
1295 			PROC_SLOCK(p);
1296 		}
1297 	} else if (mode == SINGLE_BOUNDARY) {
1298 		/*
1299 		 * Wait until all suspended threads are removed from
1300 		 * the processors.  The thread_suspend_check()
1301 		 * increments p_boundary_count while it is still
1302 		 * running, which makes it possible for the execve()
1303 		 * to destroy vmspace while our other threads are
1304 		 * still using the address space.
1305 		 *
1306 		 * We lock the thread, which is only allowed to
1307 		 * succeed after context switch code finished using
1308 		 * the address space.
1309 		 */
1310 		FOREACH_THREAD_IN_PROC(p, td2) {
1311 			if (td2 == td)
1312 				continue;
1313 			thread_lock(td2);
1314 			KASSERT((td2->td_flags & TDF_BOUNDARY) != 0,
1315 			    ("td %p not on boundary", td2));
1316 			KASSERT(TD_IS_SUSPENDED(td2),
1317 			    ("td %p is not suspended", td2));
1318 			thread_unlock(td2);
1319 		}
1320 	}
1321 	PROC_SUNLOCK(p);
1322 	return (0);
1323 }
1324 
1325 bool
1326 thread_suspend_check_needed(void)
1327 {
1328 	struct proc *p;
1329 	struct thread *td;
1330 
1331 	td = curthread;
1332 	p = td->td_proc;
1333 	PROC_LOCK_ASSERT(p, MA_OWNED);
1334 	return (P_SHOULDSTOP(p) || ((p->p_flag & P_TRACED) != 0 &&
1335 	    (td->td_dbgflags & TDB_SUSPEND) != 0));
1336 }
1337 
1338 /*
1339  * Called in from locations that can safely check to see
1340  * whether we have to suspend or at least throttle for a
1341  * single-thread event (e.g. fork).
1342  *
1343  * Such locations include userret().
1344  * If the "return_instead" argument is non zero, the thread must be able to
1345  * accept 0 (caller may continue), or 1 (caller must abort) as a result.
1346  *
1347  * The 'return_instead' argument tells the function if it may do a
1348  * thread_exit() or suspend, or whether the caller must abort and back
1349  * out instead.
1350  *
1351  * If the thread that set the single_threading request has set the
1352  * P_SINGLE_EXIT bit in the process flags then this call will never return
1353  * if 'return_instead' is false, but will exit.
1354  *
1355  * P_SINGLE_EXIT | return_instead == 0| return_instead != 0
1356  *---------------+--------------------+---------------------
1357  *       0       | returns 0          |   returns 0 or 1
1358  *               | when ST ends       |   immediately
1359  *---------------+--------------------+---------------------
1360  *       1       | thread exits       |   returns 1
1361  *               |                    |  immediately
1362  * 0 = thread_exit() or suspension ok,
1363  * other = return error instead of stopping the thread.
1364  *
1365  * While a full suspension is under effect, even a single threading
1366  * thread would be suspended if it made this call (but it shouldn't).
1367  * This call should only be made from places where
1368  * thread_exit() would be safe as that may be the outcome unless
1369  * return_instead is set.
1370  */
1371 int
1372 thread_suspend_check(int return_instead)
1373 {
1374 	struct thread *td;
1375 	struct proc *p;
1376 	int wakeup_swapper;
1377 
1378 	td = curthread;
1379 	p = td->td_proc;
1380 	mtx_assert(&Giant, MA_NOTOWNED);
1381 	PROC_LOCK_ASSERT(p, MA_OWNED);
1382 	while (thread_suspend_check_needed()) {
1383 		if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
1384 			KASSERT(p->p_singlethread != NULL,
1385 			    ("singlethread not set"));
1386 			/*
1387 			 * The only suspension in action is a
1388 			 * single-threading. Single threader need not stop.
1389 			 * It is safe to access p->p_singlethread unlocked
1390 			 * because it can only be set to our address by us.
1391 			 */
1392 			if (p->p_singlethread == td)
1393 				return (0);	/* Exempt from stopping. */
1394 		}
1395 		if ((p->p_flag & P_SINGLE_EXIT) && return_instead)
1396 			return (EINTR);
1397 
1398 		/* Should we goto user boundary if we didn't come from there? */
1399 		if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE &&
1400 		    (p->p_flag & P_SINGLE_BOUNDARY) && return_instead)
1401 			return (ERESTART);
1402 
1403 		/*
1404 		 * Ignore suspend requests if they are deferred.
1405 		 */
1406 		if ((td->td_flags & TDF_SBDRY) != 0) {
1407 			KASSERT(return_instead,
1408 			    ("TDF_SBDRY set for unsafe thread_suspend_check"));
1409 			KASSERT((td->td_flags & (TDF_SEINTR | TDF_SERESTART)) !=
1410 			    (TDF_SEINTR | TDF_SERESTART),
1411 			    ("both TDF_SEINTR and TDF_SERESTART"));
1412 			return (TD_SBDRY_INTR(td) ? TD_SBDRY_ERRNO(td) : 0);
1413 		}
1414 
1415 		/*
1416 		 * If the process is waiting for us to exit,
1417 		 * this thread should just suicide.
1418 		 * Assumes that P_SINGLE_EXIT implies P_STOPPED_SINGLE.
1419 		 */
1420 		if ((p->p_flag & P_SINGLE_EXIT) && (p->p_singlethread != td)) {
1421 			PROC_UNLOCK(p);
1422 
1423 			/*
1424 			 * Allow Linux emulation layer to do some work
1425 			 * before thread suicide.
1426 			 */
1427 			if (__predict_false(p->p_sysent->sv_thread_detach != NULL))
1428 				(p->p_sysent->sv_thread_detach)(td);
1429 			umtx_thread_exit(td);
1430 			kern_thr_exit(td);
1431 			panic("stopped thread did not exit");
1432 		}
1433 
1434 		PROC_SLOCK(p);
1435 		thread_stopped(p);
1436 		if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
1437 			if (p->p_numthreads == p->p_suspcount + 1) {
1438 				thread_lock(p->p_singlethread);
1439 				wakeup_swapper = thread_unsuspend_one(
1440 				    p->p_singlethread, p, false);
1441 				if (wakeup_swapper)
1442 					kick_proc0();
1443 			}
1444 		}
1445 		PROC_UNLOCK(p);
1446 		thread_lock(td);
1447 		/*
1448 		 * When a thread suspends, it just
1449 		 * gets taken off all queues.
1450 		 */
1451 		thread_suspend_one(td);
1452 		if (return_instead == 0) {
1453 			p->p_boundary_count++;
1454 			td->td_flags |= TDF_BOUNDARY;
1455 		}
1456 		PROC_SUNLOCK(p);
1457 		mi_switch(SW_INVOL | SWT_SUSPEND);
1458 		PROC_LOCK(p);
1459 	}
1460 	return (0);
1461 }
1462 
1463 /*
1464  * Check for possible stops and suspensions while executing a
1465  * casueword or similar transiently failing operation.
1466  *
1467  * The sleep argument controls whether the function can handle a stop
1468  * request itself or it should return ERESTART and the request is
1469  * proceed at the kernel/user boundary in ast.
1470  *
1471  * Typically, when retrying due to casueword(9) failure (rv == 1), we
1472  * should handle the stop requests there, with exception of cases when
1473  * the thread owns a kernel resource, for instance busied the umtx
1474  * key, or when functions return immediately if thread_check_susp()
1475  * returned non-zero.  On the other hand, retrying the whole lock
1476  * operation, we better not stop there but delegate the handling to
1477  * ast.
1478  *
1479  * If the request is for thread termination P_SINGLE_EXIT, we cannot
1480  * handle it at all, and simply return EINTR.
1481  */
1482 int
1483 thread_check_susp(struct thread *td, bool sleep)
1484 {
1485 	struct proc *p;
1486 	int error;
1487 
1488 	/*
1489 	 * The check for TDA_SUSPEND is racy, but it is enough to
1490 	 * eventually break the lockstep loop.
1491 	 */
1492 	if (!td_ast_pending(td, TDA_SUSPEND))
1493 		return (0);
1494 	error = 0;
1495 	p = td->td_proc;
1496 	PROC_LOCK(p);
1497 	if (p->p_flag & P_SINGLE_EXIT)
1498 		error = EINTR;
1499 	else if (P_SHOULDSTOP(p) ||
1500 	    ((p->p_flag & P_TRACED) && (td->td_dbgflags & TDB_SUSPEND)))
1501 		error = sleep ? thread_suspend_check(0) : ERESTART;
1502 	PROC_UNLOCK(p);
1503 	return (error);
1504 }
1505 
1506 void
1507 thread_suspend_switch(struct thread *td, struct proc *p)
1508 {
1509 
1510 	KASSERT(!TD_IS_SUSPENDED(td), ("already suspended"));
1511 	PROC_LOCK_ASSERT(p, MA_OWNED);
1512 	PROC_SLOCK_ASSERT(p, MA_OWNED);
1513 	/*
1514 	 * We implement thread_suspend_one in stages here to avoid
1515 	 * dropping the proc lock while the thread lock is owned.
1516 	 */
1517 	if (p == td->td_proc) {
1518 		thread_stopped(p);
1519 		p->p_suspcount++;
1520 	}
1521 	PROC_UNLOCK(p);
1522 	thread_lock(td);
1523 	ast_unsched_locked(td, TDA_SUSPEND);
1524 	TD_SET_SUSPENDED(td);
1525 	sched_sleep(td, 0);
1526 	PROC_SUNLOCK(p);
1527 	DROP_GIANT();
1528 	mi_switch(SW_VOL | SWT_SUSPEND);
1529 	PICKUP_GIANT();
1530 	PROC_LOCK(p);
1531 	PROC_SLOCK(p);
1532 }
1533 
1534 void
1535 thread_suspend_one(struct thread *td)
1536 {
1537 	struct proc *p;
1538 
1539 	p = td->td_proc;
1540 	PROC_SLOCK_ASSERT(p, MA_OWNED);
1541 	THREAD_LOCK_ASSERT(td, MA_OWNED);
1542 	KASSERT(!TD_IS_SUSPENDED(td), ("already suspended"));
1543 	p->p_suspcount++;
1544 	ast_unsched_locked(td, TDA_SUSPEND);
1545 	TD_SET_SUSPENDED(td);
1546 	sched_sleep(td, 0);
1547 }
1548 
1549 static int
1550 thread_unsuspend_one(struct thread *td, struct proc *p, bool boundary)
1551 {
1552 
1553 	THREAD_LOCK_ASSERT(td, MA_OWNED);
1554 	KASSERT(TD_IS_SUSPENDED(td), ("Thread not suspended"));
1555 	TD_CLR_SUSPENDED(td);
1556 	td->td_flags &= ~TDF_ALLPROCSUSP;
1557 	if (td->td_proc == p) {
1558 		PROC_SLOCK_ASSERT(p, MA_OWNED);
1559 		p->p_suspcount--;
1560 		if (boundary && (td->td_flags & TDF_BOUNDARY) != 0) {
1561 			td->td_flags &= ~TDF_BOUNDARY;
1562 			p->p_boundary_count--;
1563 		}
1564 	}
1565 	return (setrunnable(td, 0));
1566 }
1567 
1568 void
1569 thread_run_flash(struct thread *td)
1570 {
1571 	struct proc *p;
1572 
1573 	p = td->td_proc;
1574 	PROC_LOCK_ASSERT(p, MA_OWNED);
1575 
1576 	if (TD_ON_SLEEPQ(td))
1577 		sleepq_remove_nested(td);
1578 	else
1579 		thread_lock(td);
1580 
1581 	THREAD_LOCK_ASSERT(td, MA_OWNED);
1582 	KASSERT(TD_IS_SUSPENDED(td), ("Thread not suspended"));
1583 
1584 	TD_CLR_SUSPENDED(td);
1585 	PROC_SLOCK(p);
1586 	MPASS(p->p_suspcount > 0);
1587 	p->p_suspcount--;
1588 	PROC_SUNLOCK(p);
1589 	if (setrunnable(td, 0))
1590 		kick_proc0();
1591 }
1592 
1593 /*
1594  * Allow all threads blocked by single threading to continue running.
1595  */
1596 void
1597 thread_unsuspend(struct proc *p)
1598 {
1599 	struct thread *td;
1600 	int wakeup_swapper;
1601 
1602 	PROC_LOCK_ASSERT(p, MA_OWNED);
1603 	PROC_SLOCK_ASSERT(p, MA_OWNED);
1604 	wakeup_swapper = 0;
1605 	if (!P_SHOULDSTOP(p)) {
1606                 FOREACH_THREAD_IN_PROC(p, td) {
1607 			thread_lock(td);
1608 			if (TD_IS_SUSPENDED(td))
1609 				wakeup_swapper |= thread_unsuspend_one(td, p,
1610 				    true);
1611 			else
1612 				thread_unlock(td);
1613 		}
1614 	} else if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE &&
1615 	    p->p_numthreads == p->p_suspcount) {
1616 		/*
1617 		 * Stopping everything also did the job for the single
1618 		 * threading request. Now we've downgraded to single-threaded,
1619 		 * let it continue.
1620 		 */
1621 		if (p->p_singlethread->td_proc == p) {
1622 			thread_lock(p->p_singlethread);
1623 			wakeup_swapper = thread_unsuspend_one(
1624 			    p->p_singlethread, p, false);
1625 		}
1626 	}
1627 	if (wakeup_swapper)
1628 		kick_proc0();
1629 }
1630 
1631 /*
1632  * End the single threading mode..
1633  */
1634 void
1635 thread_single_end(struct proc *p, int mode)
1636 {
1637 	struct thread *td;
1638 	int wakeup_swapper;
1639 
1640 	KASSERT(mode == SINGLE_EXIT || mode == SINGLE_BOUNDARY ||
1641 	    mode == SINGLE_ALLPROC || mode == SINGLE_NO_EXIT,
1642 	    ("invalid mode %d", mode));
1643 	PROC_LOCK_ASSERT(p, MA_OWNED);
1644 	KASSERT((mode == SINGLE_ALLPROC && (p->p_flag & P_TOTAL_STOP) != 0) ||
1645 	    (mode != SINGLE_ALLPROC && (p->p_flag & P_TOTAL_STOP) == 0),
1646 	    ("mode %d does not match P_TOTAL_STOP", mode));
1647 	KASSERT(mode == SINGLE_ALLPROC || p->p_singlethread == curthread,
1648 	    ("thread_single_end from other thread %p %p",
1649 	    curthread, p->p_singlethread));
1650 	KASSERT(mode != SINGLE_BOUNDARY ||
1651 	    (p->p_flag & P_SINGLE_BOUNDARY) != 0,
1652 	    ("mis-matched SINGLE_BOUNDARY flags %x", p->p_flag));
1653 	p->p_flag &= ~(P_STOPPED_SINGLE | P_SINGLE_EXIT | P_SINGLE_BOUNDARY |
1654 	    P_TOTAL_STOP);
1655 	PROC_SLOCK(p);
1656 	p->p_singlethread = NULL;
1657 	wakeup_swapper = 0;
1658 	/*
1659 	 * If there are other threads they may now run,
1660 	 * unless of course there is a blanket 'stop order'
1661 	 * on the process. The single threader must be allowed
1662 	 * to continue however as this is a bad place to stop.
1663 	 */
1664 	if (p->p_numthreads != remain_for_mode(mode) && !P_SHOULDSTOP(p)) {
1665                 FOREACH_THREAD_IN_PROC(p, td) {
1666 			thread_lock(td);
1667 			if (TD_IS_SUSPENDED(td)) {
1668 				wakeup_swapper |= thread_unsuspend_one(td, p,
1669 				    true);
1670 			} else
1671 				thread_unlock(td);
1672 		}
1673 	}
1674 	KASSERT(mode != SINGLE_BOUNDARY || p->p_boundary_count == 0,
1675 	    ("inconsistent boundary count %d", p->p_boundary_count));
1676 	PROC_SUNLOCK(p);
1677 	if (wakeup_swapper)
1678 		kick_proc0();
1679 	wakeup(&p->p_flag);
1680 }
1681 
1682 /*
1683  * Locate a thread by number and return with proc lock held.
1684  *
1685  * thread exit establishes proc -> tidhash lock ordering, but lookup
1686  * takes tidhash first and needs to return locked proc.
1687  *
1688  * The problem is worked around by relying on type-safety of both
1689  * structures and doing the work in 2 steps:
1690  * - tidhash-locked lookup which saves both thread and proc pointers
1691  * - proc-locked verification that the found thread still matches
1692  */
1693 static bool
1694 tdfind_hash(lwpid_t tid, pid_t pid, struct proc **pp, struct thread **tdp)
1695 {
1696 #define RUN_THRESH	16
1697 	struct proc *p;
1698 	struct thread *td;
1699 	int run;
1700 	bool locked;
1701 
1702 	run = 0;
1703 	rw_rlock(TIDHASHLOCK(tid));
1704 	locked = true;
1705 	LIST_FOREACH(td, TIDHASH(tid), td_hash) {
1706 		if (td->td_tid != tid) {
1707 			run++;
1708 			continue;
1709 		}
1710 		p = td->td_proc;
1711 		if (pid != -1 && p->p_pid != pid) {
1712 			td = NULL;
1713 			break;
1714 		}
1715 		if (run > RUN_THRESH) {
1716 			if (rw_try_upgrade(TIDHASHLOCK(tid))) {
1717 				LIST_REMOVE(td, td_hash);
1718 				LIST_INSERT_HEAD(TIDHASH(td->td_tid),
1719 					td, td_hash);
1720 				rw_wunlock(TIDHASHLOCK(tid));
1721 				locked = false;
1722 				break;
1723 			}
1724 		}
1725 		break;
1726 	}
1727 	if (locked)
1728 		rw_runlock(TIDHASHLOCK(tid));
1729 	if (td == NULL)
1730 		return (false);
1731 	*pp = p;
1732 	*tdp = td;
1733 	return (true);
1734 }
1735 
1736 struct thread *
1737 tdfind(lwpid_t tid, pid_t pid)
1738 {
1739 	struct proc *p;
1740 	struct thread *td;
1741 
1742 	td = curthread;
1743 	if (td->td_tid == tid) {
1744 		if (pid != -1 && td->td_proc->p_pid != pid)
1745 			return (NULL);
1746 		PROC_LOCK(td->td_proc);
1747 		return (td);
1748 	}
1749 
1750 	for (;;) {
1751 		if (!tdfind_hash(tid, pid, &p, &td))
1752 			return (NULL);
1753 		PROC_LOCK(p);
1754 		if (td->td_tid != tid) {
1755 			PROC_UNLOCK(p);
1756 			continue;
1757 		}
1758 		if (td->td_proc != p) {
1759 			PROC_UNLOCK(p);
1760 			continue;
1761 		}
1762 		if (p->p_state == PRS_NEW) {
1763 			PROC_UNLOCK(p);
1764 			return (NULL);
1765 		}
1766 		return (td);
1767 	}
1768 }
1769 
1770 void
1771 tidhash_add(struct thread *td)
1772 {
1773 	rw_wlock(TIDHASHLOCK(td->td_tid));
1774 	LIST_INSERT_HEAD(TIDHASH(td->td_tid), td, td_hash);
1775 	rw_wunlock(TIDHASHLOCK(td->td_tid));
1776 }
1777 
1778 void
1779 tidhash_remove(struct thread *td)
1780 {
1781 
1782 	rw_wlock(TIDHASHLOCK(td->td_tid));
1783 	LIST_REMOVE(td, td_hash);
1784 	rw_wunlock(TIDHASHLOCK(td->td_tid));
1785 }
1786