xref: /freebsd/sys/kern/sched_4bsd.c (revision 273c26a3c3bea87a241d6879abd4f991db180bf0)
1 /*-
2  * Copyright (c) 1982, 1986, 1990, 1991, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  * (c) UNIX System Laboratories, Inc.
5  * All or some portions of this file are derived from material licensed
6  * to the University of California by American Telephone and Telegraph
7  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8  * the permission of UNIX System Laboratories, Inc.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  */
34 
35 #include <sys/cdefs.h>
36 __FBSDID("$FreeBSD$");
37 
38 #include "opt_hwpmc_hooks.h"
39 #include "opt_sched.h"
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/cpuset.h>
44 #include <sys/kernel.h>
45 #include <sys/ktr.h>
46 #include <sys/lock.h>
47 #include <sys/kthread.h>
48 #include <sys/mutex.h>
49 #include <sys/proc.h>
50 #include <sys/resourcevar.h>
51 #include <sys/sched.h>
52 #include <sys/sdt.h>
53 #include <sys/smp.h>
54 #include <sys/sysctl.h>
55 #include <sys/sx.h>
56 #include <sys/turnstile.h>
57 #include <sys/umtx.h>
58 #include <machine/pcb.h>
59 #include <machine/smp.h>
60 
61 #ifdef HWPMC_HOOKS
62 #include <sys/pmckern.h>
63 #endif
64 
65 #ifdef KDTRACE_HOOKS
66 #include <sys/dtrace_bsd.h>
67 int				dtrace_vtime_active;
68 dtrace_vtime_switch_func_t	dtrace_vtime_switch_func;
69 #endif
70 
71 /*
72  * INVERSE_ESTCPU_WEIGHT is only suitable for statclock() frequencies in
73  * the range 100-256 Hz (approximately).
74  */
75 #define	ESTCPULIM(e) \
76     min((e), INVERSE_ESTCPU_WEIGHT * (NICE_WEIGHT * (PRIO_MAX - PRIO_MIN) - \
77     RQ_PPQ) + INVERSE_ESTCPU_WEIGHT - 1)
78 #ifdef SMP
79 #define	INVERSE_ESTCPU_WEIGHT	(8 * smp_cpus)
80 #else
81 #define	INVERSE_ESTCPU_WEIGHT	8	/* 1 / (priorities per estcpu level). */
82 #endif
83 #define	NICE_WEIGHT		1	/* Priorities per nice level. */
84 
85 #define	TS_NAME_LEN (MAXCOMLEN + sizeof(" td ") + sizeof(__XSTRING(UINT_MAX)))
86 
87 /*
88  * The schedulable entity that runs a context.
89  * This is  an extension to the thread structure and is tailored to
90  * the requirements of this scheduler.
91  * All fields are protected by the scheduler lock.
92  */
93 struct td_sched {
94 	fixpt_t		ts_pctcpu;	/* %cpu during p_swtime. */
95 	u_int		ts_estcpu;	/* Estimated cpu utilization. */
96 	int		ts_cpticks;	/* Ticks of cpu time. */
97 	int		ts_slptime;	/* Seconds !RUNNING. */
98 	int		ts_slice;	/* Remaining part of time slice. */
99 	int		ts_flags;
100 	struct runq	*ts_runq;	/* runq the thread is currently on */
101 #ifdef KTR
102 	char		ts_name[TS_NAME_LEN];
103 #endif
104 };
105 
106 /* flags kept in td_flags */
107 #define TDF_DIDRUN	TDF_SCHED0	/* thread actually ran. */
108 #define TDF_BOUND	TDF_SCHED1	/* Bound to one CPU. */
109 #define	TDF_SLICEEND	TDF_SCHED2	/* Thread time slice is over. */
110 
111 /* flags kept in ts_flags */
112 #define	TSF_AFFINITY	0x0001		/* Has a non-"full" CPU set. */
113 
114 #define SKE_RUNQ_PCPU(ts)						\
115     ((ts)->ts_runq != 0 && (ts)->ts_runq != &runq)
116 
117 #define	THREAD_CAN_SCHED(td, cpu)	\
118     CPU_ISSET((cpu), &(td)->td_cpuset->cs_mask)
119 
120 _Static_assert(sizeof(struct thread) + sizeof(struct td_sched) <=
121     sizeof(struct thread0_storage),
122     "increase struct thread0_storage.t0st_sched size");
123 
124 static struct mtx sched_lock;
125 
126 static int	realstathz = 127; /* stathz is sometimes 0 and run off of hz. */
127 static int	sched_tdcnt;	/* Total runnable threads in the system. */
128 static int	sched_slice = 12; /* Thread run time before rescheduling. */
129 
130 static void	setup_runqs(void);
131 static void	schedcpu(void);
132 static void	schedcpu_thread(void);
133 static void	sched_priority(struct thread *td, u_char prio);
134 static void	sched_setup(void *dummy);
135 static void	maybe_resched(struct thread *td);
136 static void	updatepri(struct thread *td);
137 static void	resetpriority(struct thread *td);
138 static void	resetpriority_thread(struct thread *td);
139 #ifdef SMP
140 static int	sched_pickcpu(struct thread *td);
141 static int	forward_wakeup(int cpunum);
142 static void	kick_other_cpu(int pri, int cpuid);
143 #endif
144 
145 static struct kproc_desc sched_kp = {
146         "schedcpu",
147         schedcpu_thread,
148         NULL
149 };
150 SYSINIT(schedcpu, SI_SUB_LAST, SI_ORDER_FIRST, kproc_start,
151     &sched_kp);
152 SYSINIT(sched_setup, SI_SUB_RUN_QUEUE, SI_ORDER_FIRST, sched_setup, NULL);
153 
154 static void sched_initticks(void *dummy);
155 SYSINIT(sched_initticks, SI_SUB_CLOCKS, SI_ORDER_THIRD, sched_initticks,
156     NULL);
157 
158 /*
159  * Global run queue.
160  */
161 static struct runq runq;
162 
163 #ifdef SMP
164 /*
165  * Per-CPU run queues
166  */
167 static struct runq runq_pcpu[MAXCPU];
168 long runq_length[MAXCPU];
169 
170 static cpuset_t idle_cpus_mask;
171 #endif
172 
173 struct pcpuidlestat {
174 	u_int idlecalls;
175 	u_int oldidlecalls;
176 };
177 static DPCPU_DEFINE(struct pcpuidlestat, idlestat);
178 
179 static void
180 setup_runqs(void)
181 {
182 #ifdef SMP
183 	int i;
184 
185 	for (i = 0; i < MAXCPU; ++i)
186 		runq_init(&runq_pcpu[i]);
187 #endif
188 
189 	runq_init(&runq);
190 }
191 
192 static int
193 sysctl_kern_quantum(SYSCTL_HANDLER_ARGS)
194 {
195 	int error, new_val, period;
196 
197 	period = 1000000 / realstathz;
198 	new_val = period * sched_slice;
199 	error = sysctl_handle_int(oidp, &new_val, 0, req);
200 	if (error != 0 || req->newptr == NULL)
201 		return (error);
202 	if (new_val <= 0)
203 		return (EINVAL);
204 	sched_slice = imax(1, (new_val + period / 2) / period);
205 	hogticks = imax(1, (2 * hz * sched_slice + realstathz / 2) /
206 	    realstathz);
207 	return (0);
208 }
209 
210 SYSCTL_NODE(_kern, OID_AUTO, sched, CTLFLAG_RD, 0, "Scheduler");
211 
212 SYSCTL_STRING(_kern_sched, OID_AUTO, name, CTLFLAG_RD, "4BSD", 0,
213     "Scheduler name");
214 SYSCTL_PROC(_kern_sched, OID_AUTO, quantum, CTLTYPE_INT | CTLFLAG_RW,
215     NULL, 0, sysctl_kern_quantum, "I",
216     "Quantum for timeshare threads in microseconds");
217 SYSCTL_INT(_kern_sched, OID_AUTO, slice, CTLFLAG_RW, &sched_slice, 0,
218     "Quantum for timeshare threads in stathz ticks");
219 #ifdef SMP
220 /* Enable forwarding of wakeups to all other cpus */
221 static SYSCTL_NODE(_kern_sched, OID_AUTO, ipiwakeup, CTLFLAG_RD, NULL,
222     "Kernel SMP");
223 
224 static int runq_fuzz = 1;
225 SYSCTL_INT(_kern_sched, OID_AUTO, runq_fuzz, CTLFLAG_RW, &runq_fuzz, 0, "");
226 
227 static int forward_wakeup_enabled = 1;
228 SYSCTL_INT(_kern_sched_ipiwakeup, OID_AUTO, enabled, CTLFLAG_RW,
229 	   &forward_wakeup_enabled, 0,
230 	   "Forwarding of wakeup to idle CPUs");
231 
232 static int forward_wakeups_requested = 0;
233 SYSCTL_INT(_kern_sched_ipiwakeup, OID_AUTO, requested, CTLFLAG_RD,
234 	   &forward_wakeups_requested, 0,
235 	   "Requests for Forwarding of wakeup to idle CPUs");
236 
237 static int forward_wakeups_delivered = 0;
238 SYSCTL_INT(_kern_sched_ipiwakeup, OID_AUTO, delivered, CTLFLAG_RD,
239 	   &forward_wakeups_delivered, 0,
240 	   "Completed Forwarding of wakeup to idle CPUs");
241 
242 static int forward_wakeup_use_mask = 1;
243 SYSCTL_INT(_kern_sched_ipiwakeup, OID_AUTO, usemask, CTLFLAG_RW,
244 	   &forward_wakeup_use_mask, 0,
245 	   "Use the mask of idle cpus");
246 
247 static int forward_wakeup_use_loop = 0;
248 SYSCTL_INT(_kern_sched_ipiwakeup, OID_AUTO, useloop, CTLFLAG_RW,
249 	   &forward_wakeup_use_loop, 0,
250 	   "Use a loop to find idle cpus");
251 
252 #endif
253 #if 0
254 static int sched_followon = 0;
255 SYSCTL_INT(_kern_sched, OID_AUTO, followon, CTLFLAG_RW,
256 	   &sched_followon, 0,
257 	   "allow threads to share a quantum");
258 #endif
259 
260 SDT_PROVIDER_DEFINE(sched);
261 
262 SDT_PROBE_DEFINE3(sched, , , change__pri, "struct thread *",
263     "struct proc *", "uint8_t");
264 SDT_PROBE_DEFINE3(sched, , , dequeue, "struct thread *",
265     "struct proc *", "void *");
266 SDT_PROBE_DEFINE4(sched, , , enqueue, "struct thread *",
267     "struct proc *", "void *", "int");
268 SDT_PROBE_DEFINE4(sched, , , lend__pri, "struct thread *",
269     "struct proc *", "uint8_t", "struct thread *");
270 SDT_PROBE_DEFINE2(sched, , , load__change, "int", "int");
271 SDT_PROBE_DEFINE2(sched, , , off__cpu, "struct thread *",
272     "struct proc *");
273 SDT_PROBE_DEFINE(sched, , , on__cpu);
274 SDT_PROBE_DEFINE(sched, , , remain__cpu);
275 SDT_PROBE_DEFINE2(sched, , , surrender, "struct thread *",
276     "struct proc *");
277 
278 static __inline void
279 sched_load_add(void)
280 {
281 
282 	sched_tdcnt++;
283 	KTR_COUNTER0(KTR_SCHED, "load", "global load", sched_tdcnt);
284 	SDT_PROBE2(sched, , , load__change, NOCPU, sched_tdcnt);
285 }
286 
287 static __inline void
288 sched_load_rem(void)
289 {
290 
291 	sched_tdcnt--;
292 	KTR_COUNTER0(KTR_SCHED, "load", "global load", sched_tdcnt);
293 	SDT_PROBE2(sched, , , load__change, NOCPU, sched_tdcnt);
294 }
295 /*
296  * Arrange to reschedule if necessary, taking the priorities and
297  * schedulers into account.
298  */
299 static void
300 maybe_resched(struct thread *td)
301 {
302 
303 	THREAD_LOCK_ASSERT(td, MA_OWNED);
304 	if (td->td_priority < curthread->td_priority)
305 		curthread->td_flags |= TDF_NEEDRESCHED;
306 }
307 
308 /*
309  * This function is called when a thread is about to be put on run queue
310  * because it has been made runnable or its priority has been adjusted.  It
311  * determines if the new thread should be immediately preempted to.  If so,
312  * it switches to it and eventually returns true.  If not, it returns false
313  * so that the caller may place the thread on an appropriate run queue.
314  */
315 int
316 maybe_preempt(struct thread *td)
317 {
318 #ifdef PREEMPTION
319 	struct thread *ctd;
320 	int cpri, pri;
321 
322 	/*
323 	 * The new thread should not preempt the current thread if any of the
324 	 * following conditions are true:
325 	 *
326 	 *  - The kernel is in the throes of crashing (panicstr).
327 	 *  - The current thread has a higher (numerically lower) or
328 	 *    equivalent priority.  Note that this prevents curthread from
329 	 *    trying to preempt to itself.
330 	 *  - It is too early in the boot for context switches (cold is set).
331 	 *  - The current thread has an inhibitor set or is in the process of
332 	 *    exiting.  In this case, the current thread is about to switch
333 	 *    out anyways, so there's no point in preempting.  If we did,
334 	 *    the current thread would not be properly resumed as well, so
335 	 *    just avoid that whole landmine.
336 	 *  - If the new thread's priority is not a realtime priority and
337 	 *    the current thread's priority is not an idle priority and
338 	 *    FULL_PREEMPTION is disabled.
339 	 *
340 	 * If all of these conditions are false, but the current thread is in
341 	 * a nested critical section, then we have to defer the preemption
342 	 * until we exit the critical section.  Otherwise, switch immediately
343 	 * to the new thread.
344 	 */
345 	ctd = curthread;
346 	THREAD_LOCK_ASSERT(td, MA_OWNED);
347 	KASSERT((td->td_inhibitors == 0),
348 			("maybe_preempt: trying to run inhibited thread"));
349 	pri = td->td_priority;
350 	cpri = ctd->td_priority;
351 	if (panicstr != NULL || pri >= cpri || cold /* || dumping */ ||
352 	    TD_IS_INHIBITED(ctd))
353 		return (0);
354 #ifndef FULL_PREEMPTION
355 	if (pri > PRI_MAX_ITHD && cpri < PRI_MIN_IDLE)
356 		return (0);
357 #endif
358 
359 	if (ctd->td_critnest > 1) {
360 		CTR1(KTR_PROC, "maybe_preempt: in critical section %d",
361 		    ctd->td_critnest);
362 		ctd->td_owepreempt = 1;
363 		return (0);
364 	}
365 	/*
366 	 * Thread is runnable but not yet put on system run queue.
367 	 */
368 	MPASS(ctd->td_lock == td->td_lock);
369 	MPASS(TD_ON_RUNQ(td));
370 	TD_SET_RUNNING(td);
371 	CTR3(KTR_PROC, "preempting to thread %p (pid %d, %s)\n", td,
372 	    td->td_proc->p_pid, td->td_name);
373 	mi_switch(SW_INVOL | SW_PREEMPT | SWT_PREEMPT, td);
374 	/*
375 	 * td's lock pointer may have changed.  We have to return with it
376 	 * locked.
377 	 */
378 	spinlock_enter();
379 	thread_unlock(ctd);
380 	thread_lock(td);
381 	spinlock_exit();
382 	return (1);
383 #else
384 	return (0);
385 #endif
386 }
387 
388 /*
389  * Constants for digital decay and forget:
390  *	90% of (ts_estcpu) usage in 5 * loadav time
391  *	95% of (ts_pctcpu) usage in 60 seconds (load insensitive)
392  *          Note that, as ps(1) mentions, this can let percentages
393  *          total over 100% (I've seen 137.9% for 3 processes).
394  *
395  * Note that schedclock() updates ts_estcpu and p_cpticks asynchronously.
396  *
397  * We wish to decay away 90% of ts_estcpu in (5 * loadavg) seconds.
398  * That is, the system wants to compute a value of decay such
399  * that the following for loop:
400  * 	for (i = 0; i < (5 * loadavg); i++)
401  * 		ts_estcpu *= decay;
402  * will compute
403  * 	ts_estcpu *= 0.1;
404  * for all values of loadavg:
405  *
406  * Mathematically this loop can be expressed by saying:
407  * 	decay ** (5 * loadavg) ~= .1
408  *
409  * The system computes decay as:
410  * 	decay = (2 * loadavg) / (2 * loadavg + 1)
411  *
412  * We wish to prove that the system's computation of decay
413  * will always fulfill the equation:
414  * 	decay ** (5 * loadavg) ~= .1
415  *
416  * If we compute b as:
417  * 	b = 2 * loadavg
418  * then
419  * 	decay = b / (b + 1)
420  *
421  * We now need to prove two things:
422  *	1) Given factor ** (5 * loadavg) ~= .1, prove factor == b/(b+1)
423  *	2) Given b/(b+1) ** power ~= .1, prove power == (5 * loadavg)
424  *
425  * Facts:
426  *         For x close to zero, exp(x) =~ 1 + x, since
427  *              exp(x) = 0! + x**1/1! + x**2/2! + ... .
428  *              therefore exp(-1/b) =~ 1 - (1/b) = (b-1)/b.
429  *         For x close to zero, ln(1+x) =~ x, since
430  *              ln(1+x) = x - x**2/2 + x**3/3 - ...     -1 < x < 1
431  *              therefore ln(b/(b+1)) = ln(1 - 1/(b+1)) =~ -1/(b+1).
432  *         ln(.1) =~ -2.30
433  *
434  * Proof of (1):
435  *    Solve (factor)**(power) =~ .1 given power (5*loadav):
436  *	solving for factor,
437  *      ln(factor) =~ (-2.30/5*loadav), or
438  *      factor =~ exp(-1/((5/2.30)*loadav)) =~ exp(-1/(2*loadav)) =
439  *          exp(-1/b) =~ (b-1)/b =~ b/(b+1).                    QED
440  *
441  * Proof of (2):
442  *    Solve (factor)**(power) =~ .1 given factor == (b/(b+1)):
443  *	solving for power,
444  *      power*ln(b/(b+1)) =~ -2.30, or
445  *      power =~ 2.3 * (b + 1) = 4.6*loadav + 2.3 =~ 5*loadav.  QED
446  *
447  * Actual power values for the implemented algorithm are as follows:
448  *      loadav: 1       2       3       4
449  *      power:  5.68    10.32   14.94   19.55
450  */
451 
452 /* calculations for digital decay to forget 90% of usage in 5*loadav sec */
453 #define	loadfactor(loadav)	(2 * (loadav))
454 #define	decay_cpu(loadfac, cpu)	(((loadfac) * (cpu)) / ((loadfac) + FSCALE))
455 
456 /* decay 95% of `ts_pctcpu' in 60 seconds; see CCPU_SHIFT before changing */
457 static fixpt_t	ccpu = 0.95122942450071400909 * FSCALE;	/* exp(-1/20) */
458 SYSCTL_UINT(_kern, OID_AUTO, ccpu, CTLFLAG_RD, &ccpu, 0, "");
459 
460 /*
461  * If `ccpu' is not equal to `exp(-1/20)' and you still want to use the
462  * faster/more-accurate formula, you'll have to estimate CCPU_SHIFT below
463  * and possibly adjust FSHIFT in "param.h" so that (FSHIFT >= CCPU_SHIFT).
464  *
465  * To estimate CCPU_SHIFT for exp(-1/20), the following formula was used:
466  *	1 - exp(-1/20) ~= 0.0487 ~= 0.0488 == 1 (fixed pt, *11* bits).
467  *
468  * If you don't want to bother with the faster/more-accurate formula, you
469  * can set CCPU_SHIFT to (FSHIFT + 1) which will use a slower/less-accurate
470  * (more general) method of calculating the %age of CPU used by a process.
471  */
472 #define	CCPU_SHIFT	11
473 
474 /*
475  * Recompute process priorities, every hz ticks.
476  * MP-safe, called without the Giant mutex.
477  */
478 /* ARGSUSED */
479 static void
480 schedcpu(void)
481 {
482 	register fixpt_t loadfac = loadfactor(averunnable.ldavg[0]);
483 	struct thread *td;
484 	struct proc *p;
485 	struct td_sched *ts;
486 	int awake;
487 
488 	sx_slock(&allproc_lock);
489 	FOREACH_PROC_IN_SYSTEM(p) {
490 		PROC_LOCK(p);
491 		if (p->p_state == PRS_NEW) {
492 			PROC_UNLOCK(p);
493 			continue;
494 		}
495 		FOREACH_THREAD_IN_PROC(p, td) {
496 			awake = 0;
497 			ts = td_get_sched(td);
498 			thread_lock(td);
499 			/*
500 			 * Increment sleep time (if sleeping).  We
501 			 * ignore overflow, as above.
502 			 */
503 			/*
504 			 * The td_sched slptimes are not touched in wakeup
505 			 * because the thread may not HAVE everything in
506 			 * memory? XXX I think this is out of date.
507 			 */
508 			if (TD_ON_RUNQ(td)) {
509 				awake = 1;
510 				td->td_flags &= ~TDF_DIDRUN;
511 			} else if (TD_IS_RUNNING(td)) {
512 				awake = 1;
513 				/* Do not clear TDF_DIDRUN */
514 			} else if (td->td_flags & TDF_DIDRUN) {
515 				awake = 1;
516 				td->td_flags &= ~TDF_DIDRUN;
517 			}
518 
519 			/*
520 			 * ts_pctcpu is only for ps and ttyinfo().
521 			 */
522 			ts->ts_pctcpu = (ts->ts_pctcpu * ccpu) >> FSHIFT;
523 			/*
524 			 * If the td_sched has been idle the entire second,
525 			 * stop recalculating its priority until
526 			 * it wakes up.
527 			 */
528 			if (ts->ts_cpticks != 0) {
529 #if	(FSHIFT >= CCPU_SHIFT)
530 				ts->ts_pctcpu += (realstathz == 100)
531 				    ? ((fixpt_t) ts->ts_cpticks) <<
532 				    (FSHIFT - CCPU_SHIFT) :
533 				    100 * (((fixpt_t) ts->ts_cpticks)
534 				    << (FSHIFT - CCPU_SHIFT)) / realstathz;
535 #else
536 				ts->ts_pctcpu += ((FSCALE - ccpu) *
537 				    (ts->ts_cpticks *
538 				    FSCALE / realstathz)) >> FSHIFT;
539 #endif
540 				ts->ts_cpticks = 0;
541 			}
542 			/*
543 			 * If there are ANY running threads in this process,
544 			 * then don't count it as sleeping.
545 			 * XXX: this is broken.
546 			 */
547 			if (awake) {
548 				if (ts->ts_slptime > 1) {
549 					/*
550 					 * In an ideal world, this should not
551 					 * happen, because whoever woke us
552 					 * up from the long sleep should have
553 					 * unwound the slptime and reset our
554 					 * priority before we run at the stale
555 					 * priority.  Should KASSERT at some
556 					 * point when all the cases are fixed.
557 					 */
558 					updatepri(td);
559 				}
560 				ts->ts_slptime = 0;
561 			} else
562 				ts->ts_slptime++;
563 			if (ts->ts_slptime > 1) {
564 				thread_unlock(td);
565 				continue;
566 			}
567 			ts->ts_estcpu = decay_cpu(loadfac, ts->ts_estcpu);
568 		      	resetpriority(td);
569 			resetpriority_thread(td);
570 			thread_unlock(td);
571 		}
572 		PROC_UNLOCK(p);
573 	}
574 	sx_sunlock(&allproc_lock);
575 }
576 
577 /*
578  * Main loop for a kthread that executes schedcpu once a second.
579  */
580 static void
581 schedcpu_thread(void)
582 {
583 
584 	for (;;) {
585 		schedcpu();
586 		pause("-", hz);
587 	}
588 }
589 
590 /*
591  * Recalculate the priority of a process after it has slept for a while.
592  * For all load averages >= 1 and max ts_estcpu of 255, sleeping for at
593  * least six times the loadfactor will decay ts_estcpu to zero.
594  */
595 static void
596 updatepri(struct thread *td)
597 {
598 	struct td_sched *ts;
599 	fixpt_t loadfac;
600 	unsigned int newcpu;
601 
602 	ts = td_get_sched(td);
603 	loadfac = loadfactor(averunnable.ldavg[0]);
604 	if (ts->ts_slptime > 5 * loadfac)
605 		ts->ts_estcpu = 0;
606 	else {
607 		newcpu = ts->ts_estcpu;
608 		ts->ts_slptime--;	/* was incremented in schedcpu() */
609 		while (newcpu && --ts->ts_slptime)
610 			newcpu = decay_cpu(loadfac, newcpu);
611 		ts->ts_estcpu = newcpu;
612 	}
613 }
614 
615 /*
616  * Compute the priority of a process when running in user mode.
617  * Arrange to reschedule if the resulting priority is better
618  * than that of the current process.
619  */
620 static void
621 resetpriority(struct thread *td)
622 {
623 	u_int newpriority;
624 
625 	if (td->td_pri_class != PRI_TIMESHARE)
626 		return;
627 	newpriority = PUSER +
628 	    td_get_sched(td)->ts_estcpu / INVERSE_ESTCPU_WEIGHT +
629 	    NICE_WEIGHT * (td->td_proc->p_nice - PRIO_MIN);
630 	newpriority = min(max(newpriority, PRI_MIN_TIMESHARE),
631 	    PRI_MAX_TIMESHARE);
632 	sched_user_prio(td, newpriority);
633 }
634 
635 /*
636  * Update the thread's priority when the associated process's user
637  * priority changes.
638  */
639 static void
640 resetpriority_thread(struct thread *td)
641 {
642 
643 	/* Only change threads with a time sharing user priority. */
644 	if (td->td_priority < PRI_MIN_TIMESHARE ||
645 	    td->td_priority > PRI_MAX_TIMESHARE)
646 		return;
647 
648 	/* XXX the whole needresched thing is broken, but not silly. */
649 	maybe_resched(td);
650 
651 	sched_prio(td, td->td_user_pri);
652 }
653 
654 /* ARGSUSED */
655 static void
656 sched_setup(void *dummy)
657 {
658 
659 	setup_runqs();
660 
661 	/* Account for thread0. */
662 	sched_load_add();
663 }
664 
665 /*
666  * This routine determines time constants after stathz and hz are setup.
667  */
668 static void
669 sched_initticks(void *dummy)
670 {
671 
672 	realstathz = stathz ? stathz : hz;
673 	sched_slice = realstathz / 10;	/* ~100ms */
674 	hogticks = imax(1, (2 * hz * sched_slice + realstathz / 2) /
675 	    realstathz);
676 }
677 
678 /* External interfaces start here */
679 
680 /*
681  * Very early in the boot some setup of scheduler-specific
682  * parts of proc0 and of some scheduler resources needs to be done.
683  * Called from:
684  *  proc0_init()
685  */
686 void
687 schedinit(void)
688 {
689 
690 	/*
691 	 * Set up the scheduler specific parts of thread0.
692 	 */
693 	thread0.td_lock = &sched_lock;
694 	td_get_sched(&thread0)->ts_slice = sched_slice;
695 	mtx_init(&sched_lock, "sched lock", NULL, MTX_SPIN | MTX_RECURSE);
696 }
697 
698 int
699 sched_runnable(void)
700 {
701 #ifdef SMP
702 	return runq_check(&runq) + runq_check(&runq_pcpu[PCPU_GET(cpuid)]);
703 #else
704 	return runq_check(&runq);
705 #endif
706 }
707 
708 int
709 sched_rr_interval(void)
710 {
711 
712 	/* Convert sched_slice from stathz to hz. */
713 	return (imax(1, (sched_slice * hz + realstathz / 2) / realstathz));
714 }
715 
716 /*
717  * We adjust the priority of the current process.  The priority of a
718  * process gets worse as it accumulates CPU time.  The cpu usage
719  * estimator (ts_estcpu) is increased here.  resetpriority() will
720  * compute a different priority each time ts_estcpu increases by
721  * INVERSE_ESTCPU_WEIGHT (until PRI_MAX_TIMESHARE is reached).  The
722  * cpu usage estimator ramps up quite quickly when the process is
723  * running (linearly), and decays away exponentially, at a rate which
724  * is proportionally slower when the system is busy.  The basic
725  * principle is that the system will 90% forget that the process used
726  * a lot of CPU time in 5 * loadav seconds.  This causes the system to
727  * favor processes which haven't run much recently, and to round-robin
728  * among other processes.
729  */
730 void
731 sched_clock(struct thread *td)
732 {
733 	struct pcpuidlestat *stat;
734 	struct td_sched *ts;
735 
736 	THREAD_LOCK_ASSERT(td, MA_OWNED);
737 	ts = td_get_sched(td);
738 
739 	ts->ts_cpticks++;
740 	ts->ts_estcpu = ESTCPULIM(ts->ts_estcpu + 1);
741 	if ((ts->ts_estcpu % INVERSE_ESTCPU_WEIGHT) == 0) {
742 		resetpriority(td);
743 		resetpriority_thread(td);
744 	}
745 
746 	/*
747 	 * Force a context switch if the current thread has used up a full
748 	 * time slice (default is 100ms).
749 	 */
750 	if (!TD_IS_IDLETHREAD(td) && --ts->ts_slice <= 0) {
751 		ts->ts_slice = sched_slice;
752 		td->td_flags |= TDF_NEEDRESCHED | TDF_SLICEEND;
753 	}
754 
755 	stat = DPCPU_PTR(idlestat);
756 	stat->oldidlecalls = stat->idlecalls;
757 	stat->idlecalls = 0;
758 }
759 
760 /*
761  * Charge child's scheduling CPU usage to parent.
762  */
763 void
764 sched_exit(struct proc *p, struct thread *td)
765 {
766 
767 	KTR_STATE1(KTR_SCHED, "thread", sched_tdname(td), "proc exit",
768 	    "prio:%d", td->td_priority);
769 
770 	PROC_LOCK_ASSERT(p, MA_OWNED);
771 	sched_exit_thread(FIRST_THREAD_IN_PROC(p), td);
772 }
773 
774 void
775 sched_exit_thread(struct thread *td, struct thread *child)
776 {
777 
778 	KTR_STATE1(KTR_SCHED, "thread", sched_tdname(child), "exit",
779 	    "prio:%d", child->td_priority);
780 	thread_lock(td);
781 	td_get_sched(td)->ts_estcpu = ESTCPULIM(td_get_sched(td)->ts_estcpu +
782 	    td_get_sched(child)->ts_estcpu);
783 	thread_unlock(td);
784 	thread_lock(child);
785 	if ((child->td_flags & TDF_NOLOAD) == 0)
786 		sched_load_rem();
787 	thread_unlock(child);
788 }
789 
790 void
791 sched_fork(struct thread *td, struct thread *childtd)
792 {
793 	sched_fork_thread(td, childtd);
794 }
795 
796 void
797 sched_fork_thread(struct thread *td, struct thread *childtd)
798 {
799 	struct td_sched *ts, *tsc;
800 
801 	childtd->td_oncpu = NOCPU;
802 	childtd->td_lastcpu = NOCPU;
803 	childtd->td_lock = &sched_lock;
804 	childtd->td_cpuset = cpuset_ref(td->td_cpuset);
805 	childtd->td_priority = childtd->td_base_pri;
806 	ts = td_get_sched(childtd);
807 	bzero(ts, sizeof(*ts));
808 	tsc = td_get_sched(td);
809 	ts->ts_estcpu = tsc->ts_estcpu;
810 	ts->ts_flags |= (tsc->ts_flags & TSF_AFFINITY);
811 	ts->ts_slice = 1;
812 }
813 
814 void
815 sched_nice(struct proc *p, int nice)
816 {
817 	struct thread *td;
818 
819 	PROC_LOCK_ASSERT(p, MA_OWNED);
820 	p->p_nice = nice;
821 	FOREACH_THREAD_IN_PROC(p, td) {
822 		thread_lock(td);
823 		resetpriority(td);
824 		resetpriority_thread(td);
825 		thread_unlock(td);
826 	}
827 }
828 
829 void
830 sched_class(struct thread *td, int class)
831 {
832 	THREAD_LOCK_ASSERT(td, MA_OWNED);
833 	td->td_pri_class = class;
834 }
835 
836 /*
837  * Adjust the priority of a thread.
838  */
839 static void
840 sched_priority(struct thread *td, u_char prio)
841 {
842 
843 
844 	KTR_POINT3(KTR_SCHED, "thread", sched_tdname(td), "priority change",
845 	    "prio:%d", td->td_priority, "new prio:%d", prio, KTR_ATTR_LINKED,
846 	    sched_tdname(curthread));
847 	SDT_PROBE3(sched, , , change__pri, td, td->td_proc, prio);
848 	if (td != curthread && prio > td->td_priority) {
849 		KTR_POINT3(KTR_SCHED, "thread", sched_tdname(curthread),
850 		    "lend prio", "prio:%d", td->td_priority, "new prio:%d",
851 		    prio, KTR_ATTR_LINKED, sched_tdname(td));
852 		SDT_PROBE4(sched, , , lend__pri, td, td->td_proc, prio,
853 		    curthread);
854 	}
855 	THREAD_LOCK_ASSERT(td, MA_OWNED);
856 	if (td->td_priority == prio)
857 		return;
858 	td->td_priority = prio;
859 	if (TD_ON_RUNQ(td) && td->td_rqindex != (prio / RQ_PPQ)) {
860 		sched_rem(td);
861 		sched_add(td, SRQ_BORING);
862 	}
863 }
864 
865 /*
866  * Update a thread's priority when it is lent another thread's
867  * priority.
868  */
869 void
870 sched_lend_prio(struct thread *td, u_char prio)
871 {
872 
873 	td->td_flags |= TDF_BORROWING;
874 	sched_priority(td, prio);
875 }
876 
877 /*
878  * Restore a thread's priority when priority propagation is
879  * over.  The prio argument is the minimum priority the thread
880  * needs to have to satisfy other possible priority lending
881  * requests.  If the thread's regulary priority is less
882  * important than prio the thread will keep a priority boost
883  * of prio.
884  */
885 void
886 sched_unlend_prio(struct thread *td, u_char prio)
887 {
888 	u_char base_pri;
889 
890 	if (td->td_base_pri >= PRI_MIN_TIMESHARE &&
891 	    td->td_base_pri <= PRI_MAX_TIMESHARE)
892 		base_pri = td->td_user_pri;
893 	else
894 		base_pri = td->td_base_pri;
895 	if (prio >= base_pri) {
896 		td->td_flags &= ~TDF_BORROWING;
897 		sched_prio(td, base_pri);
898 	} else
899 		sched_lend_prio(td, prio);
900 }
901 
902 void
903 sched_prio(struct thread *td, u_char prio)
904 {
905 	u_char oldprio;
906 
907 	/* First, update the base priority. */
908 	td->td_base_pri = prio;
909 
910 	/*
911 	 * If the thread is borrowing another thread's priority, don't ever
912 	 * lower the priority.
913 	 */
914 	if (td->td_flags & TDF_BORROWING && td->td_priority < prio)
915 		return;
916 
917 	/* Change the real priority. */
918 	oldprio = td->td_priority;
919 	sched_priority(td, prio);
920 
921 	/*
922 	 * If the thread is on a turnstile, then let the turnstile update
923 	 * its state.
924 	 */
925 	if (TD_ON_LOCK(td) && oldprio != prio)
926 		turnstile_adjust(td, oldprio);
927 }
928 
929 void
930 sched_user_prio(struct thread *td, u_char prio)
931 {
932 
933 	THREAD_LOCK_ASSERT(td, MA_OWNED);
934 	td->td_base_user_pri = prio;
935 	if (td->td_lend_user_pri <= prio)
936 		return;
937 	td->td_user_pri = prio;
938 }
939 
940 void
941 sched_lend_user_prio(struct thread *td, u_char prio)
942 {
943 
944 	THREAD_LOCK_ASSERT(td, MA_OWNED);
945 	td->td_lend_user_pri = prio;
946 	td->td_user_pri = min(prio, td->td_base_user_pri);
947 	if (td->td_priority > td->td_user_pri)
948 		sched_prio(td, td->td_user_pri);
949 	else if (td->td_priority != td->td_user_pri)
950 		td->td_flags |= TDF_NEEDRESCHED;
951 }
952 
953 void
954 sched_sleep(struct thread *td, int pri)
955 {
956 
957 	THREAD_LOCK_ASSERT(td, MA_OWNED);
958 	td->td_slptick = ticks;
959 	td_get_sched(td)->ts_slptime = 0;
960 	if (pri != 0 && PRI_BASE(td->td_pri_class) == PRI_TIMESHARE)
961 		sched_prio(td, pri);
962 	if (TD_IS_SUSPENDED(td) || pri >= PSOCK)
963 		td->td_flags |= TDF_CANSWAP;
964 }
965 
966 void
967 sched_switch(struct thread *td, struct thread *newtd, int flags)
968 {
969 	struct mtx *tmtx;
970 	struct td_sched *ts;
971 	struct proc *p;
972 	int preempted;
973 
974 	tmtx = NULL;
975 	ts = td_get_sched(td);
976 	p = td->td_proc;
977 
978 	THREAD_LOCK_ASSERT(td, MA_OWNED);
979 
980 	/*
981 	 * Switch to the sched lock to fix things up and pick
982 	 * a new thread.
983 	 * Block the td_lock in order to avoid breaking the critical path.
984 	 */
985 	if (td->td_lock != &sched_lock) {
986 		mtx_lock_spin(&sched_lock);
987 		tmtx = thread_lock_block(td);
988 	}
989 
990 	if ((td->td_flags & TDF_NOLOAD) == 0)
991 		sched_load_rem();
992 
993 	td->td_lastcpu = td->td_oncpu;
994 	preempted = !((td->td_flags & TDF_SLICEEND) ||
995 	    (flags & SWT_RELINQUISH));
996 	td->td_flags &= ~(TDF_NEEDRESCHED | TDF_SLICEEND);
997 	td->td_owepreempt = 0;
998 	td->td_oncpu = NOCPU;
999 
1000 	/*
1001 	 * At the last moment, if this thread is still marked RUNNING,
1002 	 * then put it back on the run queue as it has not been suspended
1003 	 * or stopped or any thing else similar.  We never put the idle
1004 	 * threads on the run queue, however.
1005 	 */
1006 	if (td->td_flags & TDF_IDLETD) {
1007 		TD_SET_CAN_RUN(td);
1008 #ifdef SMP
1009 		CPU_CLR(PCPU_GET(cpuid), &idle_cpus_mask);
1010 #endif
1011 	} else {
1012 		if (TD_IS_RUNNING(td)) {
1013 			/* Put us back on the run queue. */
1014 			sched_add(td, preempted ?
1015 			    SRQ_OURSELF|SRQ_YIELDING|SRQ_PREEMPTED :
1016 			    SRQ_OURSELF|SRQ_YIELDING);
1017 		}
1018 	}
1019 	if (newtd) {
1020 		/*
1021 		 * The thread we are about to run needs to be counted
1022 		 * as if it had been added to the run queue and selected.
1023 		 * It came from:
1024 		 * * A preemption
1025 		 * * An upcall
1026 		 * * A followon
1027 		 */
1028 		KASSERT((newtd->td_inhibitors == 0),
1029 			("trying to run inhibited thread"));
1030 		newtd->td_flags |= TDF_DIDRUN;
1031         	TD_SET_RUNNING(newtd);
1032 		if ((newtd->td_flags & TDF_NOLOAD) == 0)
1033 			sched_load_add();
1034 	} else {
1035 		newtd = choosethread();
1036 		MPASS(newtd->td_lock == &sched_lock);
1037 	}
1038 
1039 	if (td != newtd) {
1040 #ifdef	HWPMC_HOOKS
1041 		if (PMC_PROC_IS_USING_PMCS(td->td_proc))
1042 			PMC_SWITCH_CONTEXT(td, PMC_FN_CSW_OUT);
1043 #endif
1044 
1045 		SDT_PROBE2(sched, , , off__cpu, newtd, newtd->td_proc);
1046 
1047                 /* I feel sleepy */
1048 		lock_profile_release_lock(&sched_lock.lock_object);
1049 #ifdef KDTRACE_HOOKS
1050 		/*
1051 		 * If DTrace has set the active vtime enum to anything
1052 		 * other than INACTIVE (0), then it should have set the
1053 		 * function to call.
1054 		 */
1055 		if (dtrace_vtime_active)
1056 			(*dtrace_vtime_switch_func)(newtd);
1057 #endif
1058 
1059 		cpu_switch(td, newtd, tmtx != NULL ? tmtx : td->td_lock);
1060 		lock_profile_obtain_lock_success(&sched_lock.lock_object,
1061 		    0, 0, __FILE__, __LINE__);
1062 		/*
1063 		 * Where am I?  What year is it?
1064 		 * We are in the same thread that went to sleep above,
1065 		 * but any amount of time may have passed. All our context
1066 		 * will still be available as will local variables.
1067 		 * PCPU values however may have changed as we may have
1068 		 * changed CPU so don't trust cached values of them.
1069 		 * New threads will go to fork_exit() instead of here
1070 		 * so if you change things here you may need to change
1071 		 * things there too.
1072 		 *
1073 		 * If the thread above was exiting it will never wake
1074 		 * up again here, so either it has saved everything it
1075 		 * needed to, or the thread_wait() or wait() will
1076 		 * need to reap it.
1077 		 */
1078 
1079 		SDT_PROBE0(sched, , , on__cpu);
1080 #ifdef	HWPMC_HOOKS
1081 		if (PMC_PROC_IS_USING_PMCS(td->td_proc))
1082 			PMC_SWITCH_CONTEXT(td, PMC_FN_CSW_IN);
1083 #endif
1084 	} else
1085 		SDT_PROBE0(sched, , , remain__cpu);
1086 
1087 #ifdef SMP
1088 	if (td->td_flags & TDF_IDLETD)
1089 		CPU_SET(PCPU_GET(cpuid), &idle_cpus_mask);
1090 #endif
1091 	sched_lock.mtx_lock = (uintptr_t)td;
1092 	td->td_oncpu = PCPU_GET(cpuid);
1093 	MPASS(td->td_lock == &sched_lock);
1094 }
1095 
1096 void
1097 sched_wakeup(struct thread *td)
1098 {
1099 	struct td_sched *ts;
1100 
1101 	THREAD_LOCK_ASSERT(td, MA_OWNED);
1102 	ts = td_get_sched(td);
1103 	td->td_flags &= ~TDF_CANSWAP;
1104 	if (ts->ts_slptime > 1) {
1105 		updatepri(td);
1106 		resetpriority(td);
1107 	}
1108 	td->td_slptick = 0;
1109 	ts->ts_slptime = 0;
1110 	ts->ts_slice = sched_slice;
1111 	sched_add(td, SRQ_BORING);
1112 }
1113 
1114 #ifdef SMP
1115 static int
1116 forward_wakeup(int cpunum)
1117 {
1118 	struct pcpu *pc;
1119 	cpuset_t dontuse, map, map2;
1120 	u_int id, me;
1121 	int iscpuset;
1122 
1123 	mtx_assert(&sched_lock, MA_OWNED);
1124 
1125 	CTR0(KTR_RUNQ, "forward_wakeup()");
1126 
1127 	if ((!forward_wakeup_enabled) ||
1128 	     (forward_wakeup_use_mask == 0 && forward_wakeup_use_loop == 0))
1129 		return (0);
1130 	if (!smp_started || cold || panicstr)
1131 		return (0);
1132 
1133 	forward_wakeups_requested++;
1134 
1135 	/*
1136 	 * Check the idle mask we received against what we calculated
1137 	 * before in the old version.
1138 	 */
1139 	me = PCPU_GET(cpuid);
1140 
1141 	/* Don't bother if we should be doing it ourself. */
1142 	if (CPU_ISSET(me, &idle_cpus_mask) &&
1143 	    (cpunum == NOCPU || me == cpunum))
1144 		return (0);
1145 
1146 	CPU_SETOF(me, &dontuse);
1147 	CPU_OR(&dontuse, &stopped_cpus);
1148 	CPU_OR(&dontuse, &hlt_cpus_mask);
1149 	CPU_ZERO(&map2);
1150 	if (forward_wakeup_use_loop) {
1151 		STAILQ_FOREACH(pc, &cpuhead, pc_allcpu) {
1152 			id = pc->pc_cpuid;
1153 			if (!CPU_ISSET(id, &dontuse) &&
1154 			    pc->pc_curthread == pc->pc_idlethread) {
1155 				CPU_SET(id, &map2);
1156 			}
1157 		}
1158 	}
1159 
1160 	if (forward_wakeup_use_mask) {
1161 		map = idle_cpus_mask;
1162 		CPU_NAND(&map, &dontuse);
1163 
1164 		/* If they are both on, compare and use loop if different. */
1165 		if (forward_wakeup_use_loop) {
1166 			if (CPU_CMP(&map, &map2)) {
1167 				printf("map != map2, loop method preferred\n");
1168 				map = map2;
1169 			}
1170 		}
1171 	} else {
1172 		map = map2;
1173 	}
1174 
1175 	/* If we only allow a specific CPU, then mask off all the others. */
1176 	if (cpunum != NOCPU) {
1177 		KASSERT((cpunum <= mp_maxcpus),("forward_wakeup: bad cpunum."));
1178 		iscpuset = CPU_ISSET(cpunum, &map);
1179 		if (iscpuset == 0)
1180 			CPU_ZERO(&map);
1181 		else
1182 			CPU_SETOF(cpunum, &map);
1183 	}
1184 	if (!CPU_EMPTY(&map)) {
1185 		forward_wakeups_delivered++;
1186 		STAILQ_FOREACH(pc, &cpuhead, pc_allcpu) {
1187 			id = pc->pc_cpuid;
1188 			if (!CPU_ISSET(id, &map))
1189 				continue;
1190 			if (cpu_idle_wakeup(pc->pc_cpuid))
1191 				CPU_CLR(id, &map);
1192 		}
1193 		if (!CPU_EMPTY(&map))
1194 			ipi_selected(map, IPI_AST);
1195 		return (1);
1196 	}
1197 	if (cpunum == NOCPU)
1198 		printf("forward_wakeup: Idle processor not found\n");
1199 	return (0);
1200 }
1201 
1202 static void
1203 kick_other_cpu(int pri, int cpuid)
1204 {
1205 	struct pcpu *pcpu;
1206 	int cpri;
1207 
1208 	pcpu = pcpu_find(cpuid);
1209 	if (CPU_ISSET(cpuid, &idle_cpus_mask)) {
1210 		forward_wakeups_delivered++;
1211 		if (!cpu_idle_wakeup(cpuid))
1212 			ipi_cpu(cpuid, IPI_AST);
1213 		return;
1214 	}
1215 
1216 	cpri = pcpu->pc_curthread->td_priority;
1217 	if (pri >= cpri)
1218 		return;
1219 
1220 #if defined(IPI_PREEMPTION) && defined(PREEMPTION)
1221 #if !defined(FULL_PREEMPTION)
1222 	if (pri <= PRI_MAX_ITHD)
1223 #endif /* ! FULL_PREEMPTION */
1224 	{
1225 		ipi_cpu(cpuid, IPI_PREEMPT);
1226 		return;
1227 	}
1228 #endif /* defined(IPI_PREEMPTION) && defined(PREEMPTION) */
1229 
1230 	pcpu->pc_curthread->td_flags |= TDF_NEEDRESCHED;
1231 	ipi_cpu(cpuid, IPI_AST);
1232 	return;
1233 }
1234 #endif /* SMP */
1235 
1236 #ifdef SMP
1237 static int
1238 sched_pickcpu(struct thread *td)
1239 {
1240 	int best, cpu;
1241 
1242 	mtx_assert(&sched_lock, MA_OWNED);
1243 
1244 	if (td->td_lastcpu != NOCPU && THREAD_CAN_SCHED(td, td->td_lastcpu))
1245 		best = td->td_lastcpu;
1246 	else
1247 		best = NOCPU;
1248 	CPU_FOREACH(cpu) {
1249 		if (!THREAD_CAN_SCHED(td, cpu))
1250 			continue;
1251 
1252 		if (best == NOCPU)
1253 			best = cpu;
1254 		else if (runq_length[cpu] < runq_length[best])
1255 			best = cpu;
1256 	}
1257 	KASSERT(best != NOCPU, ("no valid CPUs"));
1258 
1259 	return (best);
1260 }
1261 #endif
1262 
1263 void
1264 sched_add(struct thread *td, int flags)
1265 #ifdef SMP
1266 {
1267 	cpuset_t tidlemsk;
1268 	struct td_sched *ts;
1269 	u_int cpu, cpuid;
1270 	int forwarded = 0;
1271 	int single_cpu = 0;
1272 
1273 	ts = td_get_sched(td);
1274 	THREAD_LOCK_ASSERT(td, MA_OWNED);
1275 	KASSERT((td->td_inhibitors == 0),
1276 	    ("sched_add: trying to run inhibited thread"));
1277 	KASSERT((TD_CAN_RUN(td) || TD_IS_RUNNING(td)),
1278 	    ("sched_add: bad thread state"));
1279 	KASSERT(td->td_flags & TDF_INMEM,
1280 	    ("sched_add: thread swapped out"));
1281 
1282 	KTR_STATE2(KTR_SCHED, "thread", sched_tdname(td), "runq add",
1283 	    "prio:%d", td->td_priority, KTR_ATTR_LINKED,
1284 	    sched_tdname(curthread));
1285 	KTR_POINT1(KTR_SCHED, "thread", sched_tdname(curthread), "wokeup",
1286 	    KTR_ATTR_LINKED, sched_tdname(td));
1287 	SDT_PROBE4(sched, , , enqueue, td, td->td_proc, NULL,
1288 	    flags & SRQ_PREEMPTED);
1289 
1290 
1291 	/*
1292 	 * Now that the thread is moving to the run-queue, set the lock
1293 	 * to the scheduler's lock.
1294 	 */
1295 	if (td->td_lock != &sched_lock) {
1296 		mtx_lock_spin(&sched_lock);
1297 		thread_lock_set(td, &sched_lock);
1298 	}
1299 	TD_SET_RUNQ(td);
1300 
1301 	/*
1302 	 * If SMP is started and the thread is pinned or otherwise limited to
1303 	 * a specific set of CPUs, queue the thread to a per-CPU run queue.
1304 	 * Otherwise, queue the thread to the global run queue.
1305 	 *
1306 	 * If SMP has not yet been started we must use the global run queue
1307 	 * as per-CPU state may not be initialized yet and we may crash if we
1308 	 * try to access the per-CPU run queues.
1309 	 */
1310 	if (smp_started && (td->td_pinned != 0 || td->td_flags & TDF_BOUND ||
1311 	    ts->ts_flags & TSF_AFFINITY)) {
1312 		if (td->td_pinned != 0)
1313 			cpu = td->td_lastcpu;
1314 		else if (td->td_flags & TDF_BOUND) {
1315 			/* Find CPU from bound runq. */
1316 			KASSERT(SKE_RUNQ_PCPU(ts),
1317 			    ("sched_add: bound td_sched not on cpu runq"));
1318 			cpu = ts->ts_runq - &runq_pcpu[0];
1319 		} else
1320 			/* Find a valid CPU for our cpuset */
1321 			cpu = sched_pickcpu(td);
1322 		ts->ts_runq = &runq_pcpu[cpu];
1323 		single_cpu = 1;
1324 		CTR3(KTR_RUNQ,
1325 		    "sched_add: Put td_sched:%p(td:%p) on cpu%d runq", ts, td,
1326 		    cpu);
1327 	} else {
1328 		CTR2(KTR_RUNQ,
1329 		    "sched_add: adding td_sched:%p (td:%p) to gbl runq", ts,
1330 		    td);
1331 		cpu = NOCPU;
1332 		ts->ts_runq = &runq;
1333 	}
1334 
1335 	cpuid = PCPU_GET(cpuid);
1336 	if (single_cpu && cpu != cpuid) {
1337 	        kick_other_cpu(td->td_priority, cpu);
1338 	} else {
1339 		if (!single_cpu) {
1340 			tidlemsk = idle_cpus_mask;
1341 			CPU_NAND(&tidlemsk, &hlt_cpus_mask);
1342 			CPU_CLR(cpuid, &tidlemsk);
1343 
1344 			if (!CPU_ISSET(cpuid, &idle_cpus_mask) &&
1345 			    ((flags & SRQ_INTR) == 0) &&
1346 			    !CPU_EMPTY(&tidlemsk))
1347 				forwarded = forward_wakeup(cpu);
1348 		}
1349 
1350 		if (!forwarded) {
1351 			if ((flags & SRQ_YIELDING) == 0 && maybe_preempt(td))
1352 				return;
1353 			else
1354 				maybe_resched(td);
1355 		}
1356 	}
1357 
1358 	if ((td->td_flags & TDF_NOLOAD) == 0)
1359 		sched_load_add();
1360 	runq_add(ts->ts_runq, td, flags);
1361 	if (cpu != NOCPU)
1362 		runq_length[cpu]++;
1363 }
1364 #else /* SMP */
1365 {
1366 	struct td_sched *ts;
1367 
1368 	ts = td_get_sched(td);
1369 	THREAD_LOCK_ASSERT(td, MA_OWNED);
1370 	KASSERT((td->td_inhibitors == 0),
1371 	    ("sched_add: trying to run inhibited thread"));
1372 	KASSERT((TD_CAN_RUN(td) || TD_IS_RUNNING(td)),
1373 	    ("sched_add: bad thread state"));
1374 	KASSERT(td->td_flags & TDF_INMEM,
1375 	    ("sched_add: thread swapped out"));
1376 	KTR_STATE2(KTR_SCHED, "thread", sched_tdname(td), "runq add",
1377 	    "prio:%d", td->td_priority, KTR_ATTR_LINKED,
1378 	    sched_tdname(curthread));
1379 	KTR_POINT1(KTR_SCHED, "thread", sched_tdname(curthread), "wokeup",
1380 	    KTR_ATTR_LINKED, sched_tdname(td));
1381 	SDT_PROBE4(sched, , , enqueue, td, td->td_proc, NULL,
1382 	    flags & SRQ_PREEMPTED);
1383 
1384 	/*
1385 	 * Now that the thread is moving to the run-queue, set the lock
1386 	 * to the scheduler's lock.
1387 	 */
1388 	if (td->td_lock != &sched_lock) {
1389 		mtx_lock_spin(&sched_lock);
1390 		thread_lock_set(td, &sched_lock);
1391 	}
1392 	TD_SET_RUNQ(td);
1393 	CTR2(KTR_RUNQ, "sched_add: adding td_sched:%p (td:%p) to runq", ts, td);
1394 	ts->ts_runq = &runq;
1395 
1396 	/*
1397 	 * If we are yielding (on the way out anyhow) or the thread
1398 	 * being saved is US, then don't try be smart about preemption
1399 	 * or kicking off another CPU as it won't help and may hinder.
1400 	 * In the YIEDLING case, we are about to run whoever is being
1401 	 * put in the queue anyhow, and in the OURSELF case, we are
1402 	 * putting ourself on the run queue which also only happens
1403 	 * when we are about to yield.
1404 	 */
1405 	if ((flags & SRQ_YIELDING) == 0) {
1406 		if (maybe_preempt(td))
1407 			return;
1408 	}
1409 	if ((td->td_flags & TDF_NOLOAD) == 0)
1410 		sched_load_add();
1411 	runq_add(ts->ts_runq, td, flags);
1412 	maybe_resched(td);
1413 }
1414 #endif /* SMP */
1415 
1416 void
1417 sched_rem(struct thread *td)
1418 {
1419 	struct td_sched *ts;
1420 
1421 	ts = td_get_sched(td);
1422 	KASSERT(td->td_flags & TDF_INMEM,
1423 	    ("sched_rem: thread swapped out"));
1424 	KASSERT(TD_ON_RUNQ(td),
1425 	    ("sched_rem: thread not on run queue"));
1426 	mtx_assert(&sched_lock, MA_OWNED);
1427 	KTR_STATE2(KTR_SCHED, "thread", sched_tdname(td), "runq rem",
1428 	    "prio:%d", td->td_priority, KTR_ATTR_LINKED,
1429 	    sched_tdname(curthread));
1430 	SDT_PROBE3(sched, , , dequeue, td, td->td_proc, NULL);
1431 
1432 	if ((td->td_flags & TDF_NOLOAD) == 0)
1433 		sched_load_rem();
1434 #ifdef SMP
1435 	if (ts->ts_runq != &runq)
1436 		runq_length[ts->ts_runq - runq_pcpu]--;
1437 #endif
1438 	runq_remove(ts->ts_runq, td);
1439 	TD_SET_CAN_RUN(td);
1440 }
1441 
1442 /*
1443  * Select threads to run.  Note that running threads still consume a
1444  * slot.
1445  */
1446 struct thread *
1447 sched_choose(void)
1448 {
1449 	struct thread *td;
1450 	struct runq *rq;
1451 
1452 	mtx_assert(&sched_lock,  MA_OWNED);
1453 #ifdef SMP
1454 	struct thread *tdcpu;
1455 
1456 	rq = &runq;
1457 	td = runq_choose_fuzz(&runq, runq_fuzz);
1458 	tdcpu = runq_choose(&runq_pcpu[PCPU_GET(cpuid)]);
1459 
1460 	if (td == NULL ||
1461 	    (tdcpu != NULL &&
1462 	     tdcpu->td_priority < td->td_priority)) {
1463 		CTR2(KTR_RUNQ, "choosing td %p from pcpu runq %d", tdcpu,
1464 		     PCPU_GET(cpuid));
1465 		td = tdcpu;
1466 		rq = &runq_pcpu[PCPU_GET(cpuid)];
1467 	} else {
1468 		CTR1(KTR_RUNQ, "choosing td_sched %p from main runq", td);
1469 	}
1470 
1471 #else
1472 	rq = &runq;
1473 	td = runq_choose(&runq);
1474 #endif
1475 
1476 	if (td) {
1477 #ifdef SMP
1478 		if (td == tdcpu)
1479 			runq_length[PCPU_GET(cpuid)]--;
1480 #endif
1481 		runq_remove(rq, td);
1482 		td->td_flags |= TDF_DIDRUN;
1483 
1484 		KASSERT(td->td_flags & TDF_INMEM,
1485 		    ("sched_choose: thread swapped out"));
1486 		return (td);
1487 	}
1488 	return (PCPU_GET(idlethread));
1489 }
1490 
1491 void
1492 sched_preempt(struct thread *td)
1493 {
1494 
1495 	SDT_PROBE2(sched, , , surrender, td, td->td_proc);
1496 	thread_lock(td);
1497 	if (td->td_critnest > 1)
1498 		td->td_owepreempt = 1;
1499 	else
1500 		mi_switch(SW_INVOL | SW_PREEMPT | SWT_PREEMPT, NULL);
1501 	thread_unlock(td);
1502 }
1503 
1504 void
1505 sched_userret(struct thread *td)
1506 {
1507 	/*
1508 	 * XXX we cheat slightly on the locking here to avoid locking in
1509 	 * the usual case.  Setting td_priority here is essentially an
1510 	 * incomplete workaround for not setting it properly elsewhere.
1511 	 * Now that some interrupt handlers are threads, not setting it
1512 	 * properly elsewhere can clobber it in the window between setting
1513 	 * it here and returning to user mode, so don't waste time setting
1514 	 * it perfectly here.
1515 	 */
1516 	KASSERT((td->td_flags & TDF_BORROWING) == 0,
1517 	    ("thread with borrowed priority returning to userland"));
1518 	if (td->td_priority != td->td_user_pri) {
1519 		thread_lock(td);
1520 		td->td_priority = td->td_user_pri;
1521 		td->td_base_pri = td->td_user_pri;
1522 		thread_unlock(td);
1523 	}
1524 }
1525 
1526 void
1527 sched_bind(struct thread *td, int cpu)
1528 {
1529 	struct td_sched *ts;
1530 
1531 	THREAD_LOCK_ASSERT(td, MA_OWNED|MA_NOTRECURSED);
1532 	KASSERT(td == curthread, ("sched_bind: can only bind curthread"));
1533 
1534 	ts = td_get_sched(td);
1535 
1536 	td->td_flags |= TDF_BOUND;
1537 #ifdef SMP
1538 	ts->ts_runq = &runq_pcpu[cpu];
1539 	if (PCPU_GET(cpuid) == cpu)
1540 		return;
1541 
1542 	mi_switch(SW_VOL, NULL);
1543 #endif
1544 }
1545 
1546 void
1547 sched_unbind(struct thread* td)
1548 {
1549 	THREAD_LOCK_ASSERT(td, MA_OWNED);
1550 	KASSERT(td == curthread, ("sched_unbind: can only bind curthread"));
1551 	td->td_flags &= ~TDF_BOUND;
1552 }
1553 
1554 int
1555 sched_is_bound(struct thread *td)
1556 {
1557 	THREAD_LOCK_ASSERT(td, MA_OWNED);
1558 	return (td->td_flags & TDF_BOUND);
1559 }
1560 
1561 void
1562 sched_relinquish(struct thread *td)
1563 {
1564 	thread_lock(td);
1565 	mi_switch(SW_VOL | SWT_RELINQUISH, NULL);
1566 	thread_unlock(td);
1567 }
1568 
1569 int
1570 sched_load(void)
1571 {
1572 	return (sched_tdcnt);
1573 }
1574 
1575 int
1576 sched_sizeof_proc(void)
1577 {
1578 	return (sizeof(struct proc));
1579 }
1580 
1581 int
1582 sched_sizeof_thread(void)
1583 {
1584 	return (sizeof(struct thread) + sizeof(struct td_sched));
1585 }
1586 
1587 fixpt_t
1588 sched_pctcpu(struct thread *td)
1589 {
1590 	struct td_sched *ts;
1591 
1592 	THREAD_LOCK_ASSERT(td, MA_OWNED);
1593 	ts = td_get_sched(td);
1594 	return (ts->ts_pctcpu);
1595 }
1596 
1597 #ifdef RACCT
1598 /*
1599  * Calculates the contribution to the thread cpu usage for the latest
1600  * (unfinished) second.
1601  */
1602 fixpt_t
1603 sched_pctcpu_delta(struct thread *td)
1604 {
1605 	struct td_sched *ts;
1606 	fixpt_t delta;
1607 	int realstathz;
1608 
1609 	THREAD_LOCK_ASSERT(td, MA_OWNED);
1610 	ts = td_get_sched(td);
1611 	delta = 0;
1612 	realstathz = stathz ? stathz : hz;
1613 	if (ts->ts_cpticks != 0) {
1614 #if	(FSHIFT >= CCPU_SHIFT)
1615 		delta = (realstathz == 100)
1616 		    ? ((fixpt_t) ts->ts_cpticks) <<
1617 		    (FSHIFT - CCPU_SHIFT) :
1618 		    100 * (((fixpt_t) ts->ts_cpticks)
1619 		    << (FSHIFT - CCPU_SHIFT)) / realstathz;
1620 #else
1621 		delta = ((FSCALE - ccpu) *
1622 		    (ts->ts_cpticks *
1623 		    FSCALE / realstathz)) >> FSHIFT;
1624 #endif
1625 	}
1626 
1627 	return (delta);
1628 }
1629 #endif
1630 
1631 u_int
1632 sched_estcpu(struct thread *td)
1633 {
1634 
1635 	return (td_get_sched(td)->ts_estcpu);
1636 }
1637 
1638 /*
1639  * The actual idle process.
1640  */
1641 void
1642 sched_idletd(void *dummy)
1643 {
1644 	struct pcpuidlestat *stat;
1645 
1646 	THREAD_NO_SLEEPING();
1647 	stat = DPCPU_PTR(idlestat);
1648 	for (;;) {
1649 		mtx_assert(&Giant, MA_NOTOWNED);
1650 
1651 		while (sched_runnable() == 0) {
1652 			cpu_idle(stat->idlecalls + stat->oldidlecalls > 64);
1653 			stat->idlecalls++;
1654 		}
1655 
1656 		mtx_lock_spin(&sched_lock);
1657 		mi_switch(SW_VOL | SWT_IDLE, NULL);
1658 		mtx_unlock_spin(&sched_lock);
1659 	}
1660 }
1661 
1662 /*
1663  * A CPU is entering for the first time or a thread is exiting.
1664  */
1665 void
1666 sched_throw(struct thread *td)
1667 {
1668 	/*
1669 	 * Correct spinlock nesting.  The idle thread context that we are
1670 	 * borrowing was created so that it would start out with a single
1671 	 * spin lock (sched_lock) held in fork_trampoline().  Since we've
1672 	 * explicitly acquired locks in this function, the nesting count
1673 	 * is now 2 rather than 1.  Since we are nested, calling
1674 	 * spinlock_exit() will simply adjust the counts without allowing
1675 	 * spin lock using code to interrupt us.
1676 	 */
1677 	if (td == NULL) {
1678 		mtx_lock_spin(&sched_lock);
1679 		spinlock_exit();
1680 		PCPU_SET(switchtime, cpu_ticks());
1681 		PCPU_SET(switchticks, ticks);
1682 	} else {
1683 		lock_profile_release_lock(&sched_lock.lock_object);
1684 		MPASS(td->td_lock == &sched_lock);
1685 		td->td_lastcpu = td->td_oncpu;
1686 		td->td_oncpu = NOCPU;
1687 	}
1688 	mtx_assert(&sched_lock, MA_OWNED);
1689 	KASSERT(curthread->td_md.md_spinlock_count == 1, ("invalid count"));
1690 	cpu_throw(td, choosethread());	/* doesn't return */
1691 }
1692 
1693 void
1694 sched_fork_exit(struct thread *td)
1695 {
1696 
1697 	/*
1698 	 * Finish setting up thread glue so that it begins execution in a
1699 	 * non-nested critical section with sched_lock held but not recursed.
1700 	 */
1701 	td->td_oncpu = PCPU_GET(cpuid);
1702 	sched_lock.mtx_lock = (uintptr_t)td;
1703 	lock_profile_obtain_lock_success(&sched_lock.lock_object,
1704 	    0, 0, __FILE__, __LINE__);
1705 	THREAD_LOCK_ASSERT(td, MA_OWNED | MA_NOTRECURSED);
1706 }
1707 
1708 char *
1709 sched_tdname(struct thread *td)
1710 {
1711 #ifdef KTR
1712 	struct td_sched *ts;
1713 
1714 	ts = td_get_sched(td);
1715 	if (ts->ts_name[0] == '\0')
1716 		snprintf(ts->ts_name, sizeof(ts->ts_name),
1717 		    "%s tid %d", td->td_name, td->td_tid);
1718 	return (ts->ts_name);
1719 #else
1720 	return (td->td_name);
1721 #endif
1722 }
1723 
1724 #ifdef KTR
1725 void
1726 sched_clear_tdname(struct thread *td)
1727 {
1728 	struct td_sched *ts;
1729 
1730 	ts = td_get_sched(td);
1731 	ts->ts_name[0] = '\0';
1732 }
1733 #endif
1734 
1735 void
1736 sched_affinity(struct thread *td)
1737 {
1738 #ifdef SMP
1739 	struct td_sched *ts;
1740 	int cpu;
1741 
1742 	THREAD_LOCK_ASSERT(td, MA_OWNED);
1743 
1744 	/*
1745 	 * Set the TSF_AFFINITY flag if there is at least one CPU this
1746 	 * thread can't run on.
1747 	 */
1748 	ts = td_get_sched(td);
1749 	ts->ts_flags &= ~TSF_AFFINITY;
1750 	CPU_FOREACH(cpu) {
1751 		if (!THREAD_CAN_SCHED(td, cpu)) {
1752 			ts->ts_flags |= TSF_AFFINITY;
1753 			break;
1754 		}
1755 	}
1756 
1757 	/*
1758 	 * If this thread can run on all CPUs, nothing else to do.
1759 	 */
1760 	if (!(ts->ts_flags & TSF_AFFINITY))
1761 		return;
1762 
1763 	/* Pinned threads and bound threads should be left alone. */
1764 	if (td->td_pinned != 0 || td->td_flags & TDF_BOUND)
1765 		return;
1766 
1767 	switch (td->td_state) {
1768 	case TDS_RUNQ:
1769 		/*
1770 		 * If we are on a per-CPU runqueue that is in the set,
1771 		 * then nothing needs to be done.
1772 		 */
1773 		if (ts->ts_runq != &runq &&
1774 		    THREAD_CAN_SCHED(td, ts->ts_runq - runq_pcpu))
1775 			return;
1776 
1777 		/* Put this thread on a valid per-CPU runqueue. */
1778 		sched_rem(td);
1779 		sched_add(td, SRQ_BORING);
1780 		break;
1781 	case TDS_RUNNING:
1782 		/*
1783 		 * See if our current CPU is in the set.  If not, force a
1784 		 * context switch.
1785 		 */
1786 		if (THREAD_CAN_SCHED(td, td->td_oncpu))
1787 			return;
1788 
1789 		td->td_flags |= TDF_NEEDRESCHED;
1790 		if (td != curthread)
1791 			ipi_cpu(cpu, IPI_AST);
1792 		break;
1793 	default:
1794 		break;
1795 	}
1796 #endif
1797 }
1798