xref: /freebsd/sys/kern/sched_ule.c (revision dce6e6518b85561495cff38a3074a69d29d58a55)
1 /*-
2  * Copyright (c) 2002-2003, Jeffrey Roberson <jeff@freebsd.org>
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice unmodified, this list of conditions, and the following
10  *    disclaimer.
11  * 2. Redistributions in binary form must reproduce the above copyright
12  *    notice, this list of conditions and the following disclaimer in the
13  *    documentation and/or other materials provided with the distribution.
14  *
15  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
16  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
17  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
18  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
19  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
20  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
21  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
22  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
23  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
24  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
25  */
26 
27 #include <sys/cdefs.h>
28 __FBSDID("$FreeBSD$");
29 
30 #include <sys/param.h>
31 #include <sys/systm.h>
32 #include <sys/kernel.h>
33 #include <sys/ktr.h>
34 #include <sys/lock.h>
35 #include <sys/mutex.h>
36 #include <sys/proc.h>
37 #include <sys/resource.h>
38 #include <sys/sched.h>
39 #include <sys/smp.h>
40 #include <sys/sx.h>
41 #include <sys/sysctl.h>
42 #include <sys/sysproto.h>
43 #include <sys/vmmeter.h>
44 #ifdef DDB
45 #include <ddb/ddb.h>
46 #endif
47 #ifdef KTRACE
48 #include <sys/uio.h>
49 #include <sys/ktrace.h>
50 #endif
51 
52 #include <machine/cpu.h>
53 
54 #define KTR_ULE         KTR_NFS
55 
56 /* decay 95% of `p_pctcpu' in 60 seconds; see CCPU_SHIFT before changing */
57 /* XXX This is bogus compatability crap for ps */
58 static fixpt_t  ccpu = 0.95122942450071400909 * FSCALE; /* exp(-1/20) */
59 SYSCTL_INT(_kern, OID_AUTO, ccpu, CTLFLAG_RD, &ccpu, 0, "");
60 
61 static void sched_setup(void *dummy);
62 SYSINIT(sched_setup, SI_SUB_RUN_QUEUE, SI_ORDER_FIRST, sched_setup, NULL)
63 
64 static SYSCTL_NODE(_kern, OID_AUTO, sched, CTLFLAG_RW, 0, "SCHED");
65 
66 static int sched_strict;
67 SYSCTL_INT(_kern_sched, OID_AUTO, strict, CTLFLAG_RD, &sched_strict, 0, "");
68 
69 static int slice_min = 1;
70 SYSCTL_INT(_kern_sched, OID_AUTO, slice_min, CTLFLAG_RW, &slice_min, 0, "");
71 
72 static int slice_max = 10;
73 SYSCTL_INT(_kern_sched, OID_AUTO, slice_max, CTLFLAG_RW, &slice_max, 0, "");
74 
75 int realstathz;
76 int tickincr = 1;
77 
78 #ifdef SMP
79 /* Callout to handle load balancing SMP systems. */
80 static struct callout kseq_lb_callout;
81 #endif
82 
83 /*
84  * These datastructures are allocated within their parent datastructure but
85  * are scheduler specific.
86  */
87 
88 struct ke_sched {
89 	int		ske_slice;
90 	struct runq	*ske_runq;
91 	/* The following variables are only used for pctcpu calculation */
92 	int		ske_ltick;	/* Last tick that we were running on */
93 	int		ske_ftick;	/* First tick that we were running on */
94 	int		ske_ticks;	/* Tick count */
95 	/* CPU that we have affinity for. */
96 	u_char		ske_cpu;
97 };
98 #define	ke_slice	ke_sched->ske_slice
99 #define	ke_runq		ke_sched->ske_runq
100 #define	ke_ltick	ke_sched->ske_ltick
101 #define	ke_ftick	ke_sched->ske_ftick
102 #define	ke_ticks	ke_sched->ske_ticks
103 #define	ke_cpu		ke_sched->ske_cpu
104 
105 struct kg_sched {
106 	int	skg_slptime;		/* Number of ticks we vol. slept */
107 	int	skg_runtime;		/* Number of ticks we were running */
108 };
109 #define	kg_slptime	kg_sched->skg_slptime
110 #define	kg_runtime	kg_sched->skg_runtime
111 
112 struct td_sched {
113 	int	std_slptime;
114 };
115 #define	td_slptime	td_sched->std_slptime
116 
117 struct td_sched td_sched;
118 struct ke_sched ke_sched;
119 struct kg_sched kg_sched;
120 
121 struct ke_sched *kse0_sched = &ke_sched;
122 struct kg_sched *ksegrp0_sched = &kg_sched;
123 struct p_sched *proc0_sched = NULL;
124 struct td_sched *thread0_sched = &td_sched;
125 
126 /*
127  * The priority is primarily determined by the interactivity score.  Thus, we
128  * give lower(better) priorities to kse groups that use less CPU.  The nice
129  * value is then directly added to this to allow nice to have some effect
130  * on latency.
131  *
132  * PRI_RANGE:	Total priority range for timeshare threads.
133  * PRI_NRESV:	Number of nice values.
134  * PRI_BASE:	The start of the dynamic range.
135  */
136 #define	SCHED_PRI_RANGE		(PRI_MAX_TIMESHARE - PRI_MIN_TIMESHARE + 1)
137 #define	SCHED_PRI_NRESV		PRIO_TOTAL
138 #define	SCHED_PRI_NHALF		(PRIO_TOTAL / 2)
139 #define	SCHED_PRI_NTHRESH	(SCHED_PRI_NHALF - 1)
140 #define	SCHED_PRI_BASE		(PRI_MIN_TIMESHARE)
141 #define	SCHED_PRI_INTERACT(score)					\
142     ((score) * SCHED_PRI_RANGE / SCHED_INTERACT_MAX)
143 
144 /*
145  * These determine the interactivity of a process.
146  *
147  * SLP_RUN_MAX:	Maximum amount of sleep time + run time we'll accumulate
148  *		before throttling back.
149  * SLP_RUN_THROTTLE:	Divisor for reducing slp/run time at fork time.
150  * INTERACT_MAX:	Maximum interactivity value.  Smaller is better.
151  * INTERACT_THRESH:	Threshhold for placement on the current runq.
152  */
153 #define	SCHED_SLP_RUN_MAX	((hz * 2) << 10)
154 #define	SCHED_SLP_RUN_THROTTLE	(100)
155 #define	SCHED_INTERACT_MAX	(100)
156 #define	SCHED_INTERACT_HALF	(SCHED_INTERACT_MAX / 2)
157 #define	SCHED_INTERACT_THRESH	(20)
158 
159 /*
160  * These parameters and macros determine the size of the time slice that is
161  * granted to each thread.
162  *
163  * SLICE_MIN:	Minimum time slice granted, in units of ticks.
164  * SLICE_MAX:	Maximum time slice granted.
165  * SLICE_RANGE:	Range of available time slices scaled by hz.
166  * SLICE_SCALE:	The number slices granted per val in the range of [0, max].
167  * SLICE_NICE:  Determine the amount of slice granted to a scaled nice.
168  */
169 #define	SCHED_SLICE_MIN			(slice_min)
170 #define	SCHED_SLICE_MAX			(slice_max)
171 #define	SCHED_SLICE_RANGE		(SCHED_SLICE_MAX - SCHED_SLICE_MIN + 1)
172 #define	SCHED_SLICE_SCALE(val, max)	(((val) * SCHED_SLICE_RANGE) / (max))
173 #define	SCHED_SLICE_NICE(nice)						\
174     (SCHED_SLICE_MAX - SCHED_SLICE_SCALE((nice), SCHED_PRI_NTHRESH))
175 
176 /*
177  * This macro determines whether or not the kse belongs on the current or
178  * next run queue.
179  *
180  * XXX nice value should effect how interactive a kg is.
181  */
182 #define	SCHED_INTERACTIVE(kg)						\
183     (sched_interact_score(kg) < SCHED_INTERACT_THRESH)
184 #define	SCHED_CURR(kg, ke)						\
185     (ke->ke_thread->td_priority < PRI_MIN_TIMESHARE || SCHED_INTERACTIVE(kg))
186 
187 /*
188  * Cpu percentage computation macros and defines.
189  *
190  * SCHED_CPU_TIME:	Number of seconds to average the cpu usage across.
191  * SCHED_CPU_TICKS:	Number of hz ticks to average the cpu usage across.
192  */
193 
194 #define	SCHED_CPU_TIME	10
195 #define	SCHED_CPU_TICKS	(hz * SCHED_CPU_TIME)
196 
197 /*
198  * kseq - per processor runqs and statistics.
199  */
200 
201 #define	KSEQ_NCLASS	(PRI_IDLE + 1)	/* Number of run classes. */
202 
203 struct kseq {
204 	struct runq	ksq_idle;		/* Queue of IDLE threads. */
205 	struct runq	ksq_timeshare[2];	/* Run queues for !IDLE. */
206 	struct runq	*ksq_next;		/* Next timeshare queue. */
207 	struct runq	*ksq_curr;		/* Current queue. */
208 	int		ksq_loads[KSEQ_NCLASS];	/* Load for each class */
209 	int		ksq_load;		/* Aggregate load. */
210 	short		ksq_nice[PRIO_TOTAL + 1]; /* KSEs in each nice bin. */
211 	short		ksq_nicemin;		/* Least nice. */
212 #ifdef SMP
213 	int		ksq_cpus;	/* Count of CPUs in this kseq. */
214 	unsigned int	ksq_rslices;	/* Slices on run queue */
215 #endif
216 };
217 
218 /*
219  * One kse queue per processor.
220  */
221 #ifdef SMP
222 struct kseq	kseq_cpu[MAXCPU];
223 struct kseq	*kseq_idmap[MAXCPU];
224 #define	KSEQ_SELF()	(kseq_idmap[PCPU_GET(cpuid)])
225 #define	KSEQ_CPU(x)	(kseq_idmap[(x)])
226 #else
227 struct kseq	kseq_cpu;
228 #define	KSEQ_SELF()	(&kseq_cpu)
229 #define	KSEQ_CPU(x)	(&kseq_cpu)
230 #endif
231 
232 static void sched_slice(struct kse *ke);
233 static void sched_priority(struct ksegrp *kg);
234 static int sched_interact_score(struct ksegrp *kg);
235 static void sched_interact_update(struct ksegrp *kg);
236 void sched_pctcpu_update(struct kse *ke);
237 int sched_pickcpu(void);
238 
239 /* Operations on per processor queues */
240 static struct kse * kseq_choose(struct kseq *kseq);
241 static void kseq_setup(struct kseq *kseq);
242 static void kseq_add(struct kseq *kseq, struct kse *ke);
243 static void kseq_rem(struct kseq *kseq, struct kse *ke);
244 static void kseq_nice_add(struct kseq *kseq, int nice);
245 static void kseq_nice_rem(struct kseq *kseq, int nice);
246 void kseq_print(int cpu);
247 #ifdef SMP
248 struct kseq * kseq_load_highest(void);
249 void kseq_balance(void *arg);
250 void kseq_move(struct kseq *from, int cpu);
251 #endif
252 
253 void
254 kseq_print(int cpu)
255 {
256 	struct kseq *kseq;
257 	int i;
258 
259 	kseq = KSEQ_CPU(cpu);
260 
261 	printf("kseq:\n");
262 	printf("\tload:           %d\n", kseq->ksq_load);
263 	printf("\tload ITHD:      %d\n", kseq->ksq_loads[PRI_ITHD]);
264 	printf("\tload REALTIME:  %d\n", kseq->ksq_loads[PRI_REALTIME]);
265 	printf("\tload TIMESHARE: %d\n", kseq->ksq_loads[PRI_TIMESHARE]);
266 	printf("\tload IDLE:      %d\n", kseq->ksq_loads[PRI_IDLE]);
267 	printf("\tnicemin:\t%d\n", kseq->ksq_nicemin);
268 	printf("\tnice counts:\n");
269 	for (i = 0; i < PRIO_TOTAL + 1; i++)
270 		if (kseq->ksq_nice[i])
271 			printf("\t\t%d = %d\n",
272 			    i - SCHED_PRI_NHALF, kseq->ksq_nice[i]);
273 }
274 
275 static void
276 kseq_add(struct kseq *kseq, struct kse *ke)
277 {
278 	mtx_assert(&sched_lock, MA_OWNED);
279 	kseq->ksq_loads[PRI_BASE(ke->ke_ksegrp->kg_pri_class)]++;
280 	kseq->ksq_load++;
281 	if (ke->ke_ksegrp->kg_pri_class == PRI_TIMESHARE)
282 	CTR6(KTR_ULE, "Add kse %p to %p (slice: %d, pri: %d, nice: %d(%d))",
283 	    ke, ke->ke_runq, ke->ke_slice, ke->ke_thread->td_priority,
284 	    ke->ke_ksegrp->kg_nice, kseq->ksq_nicemin);
285 	if (ke->ke_ksegrp->kg_pri_class == PRI_TIMESHARE)
286 		kseq_nice_add(kseq, ke->ke_ksegrp->kg_nice);
287 #ifdef SMP
288 	kseq->ksq_rslices += ke->ke_slice;
289 #endif
290 }
291 
292 static void
293 kseq_rem(struct kseq *kseq, struct kse *ke)
294 {
295 	mtx_assert(&sched_lock, MA_OWNED);
296 	kseq->ksq_loads[PRI_BASE(ke->ke_ksegrp->kg_pri_class)]--;
297 	kseq->ksq_load--;
298 	ke->ke_runq = NULL;
299 	if (ke->ke_ksegrp->kg_pri_class == PRI_TIMESHARE)
300 		kseq_nice_rem(kseq, ke->ke_ksegrp->kg_nice);
301 #ifdef SMP
302 	kseq->ksq_rslices -= ke->ke_slice;
303 #endif
304 }
305 
306 static void
307 kseq_nice_add(struct kseq *kseq, int nice)
308 {
309 	mtx_assert(&sched_lock, MA_OWNED);
310 	/* Normalize to zero. */
311 	kseq->ksq_nice[nice + SCHED_PRI_NHALF]++;
312 	if (nice < kseq->ksq_nicemin || kseq->ksq_loads[PRI_TIMESHARE] == 1)
313 		kseq->ksq_nicemin = nice;
314 }
315 
316 static void
317 kseq_nice_rem(struct kseq *kseq, int nice)
318 {
319 	int n;
320 
321 	mtx_assert(&sched_lock, MA_OWNED);
322 	/* Normalize to zero. */
323 	n = nice + SCHED_PRI_NHALF;
324 	kseq->ksq_nice[n]--;
325 	KASSERT(kseq->ksq_nice[n] >= 0, ("Negative nice count."));
326 
327 	/*
328 	 * If this wasn't the smallest nice value or there are more in
329 	 * this bucket we can just return.  Otherwise we have to recalculate
330 	 * the smallest nice.
331 	 */
332 	if (nice != kseq->ksq_nicemin ||
333 	    kseq->ksq_nice[n] != 0 ||
334 	    kseq->ksq_loads[PRI_TIMESHARE] == 0)
335 		return;
336 
337 	for (; n < SCHED_PRI_NRESV + 1; n++)
338 		if (kseq->ksq_nice[n]) {
339 			kseq->ksq_nicemin = n - SCHED_PRI_NHALF;
340 			return;
341 		}
342 }
343 
344 #ifdef SMP
345 /*
346  * kseq_balance is a simple CPU load balancing algorithm.  It operates by
347  * finding the least loaded and most loaded cpu and equalizing their load
348  * by migrating some processes.
349  *
350  * Dealing only with two CPUs at a time has two advantages.  Firstly, most
351  * installations will only have 2 cpus.  Secondly, load balancing too much at
352  * once can have an unpleasant effect on the system.  The scheduler rarely has
353  * enough information to make perfect decisions.  So this algorithm chooses
354  * algorithm simplicity and more gradual effects on load in larger systems.
355  *
356  * It could be improved by considering the priorities and slices assigned to
357  * each task prior to balancing them.  There are many pathological cases with
358  * any approach and so the semi random algorithm below may work as well as any.
359  *
360  */
361 void
362 kseq_balance(void *arg)
363 {
364 	struct kseq *kseq;
365 	int high_load;
366 	int low_load;
367 	int high_cpu;
368 	int low_cpu;
369 	int move;
370 	int diff;
371 	int i;
372 
373 	high_cpu = 0;
374 	low_cpu = 0;
375 	high_load = 0;
376 	low_load = -1;
377 
378 	mtx_lock_spin(&sched_lock);
379 	if (smp_started == 0)
380 		goto out;
381 
382 	for (i = 0; i < mp_maxid; i++) {
383 		if (CPU_ABSENT(i) || (i & stopped_cpus) != 0)
384 			continue;
385 		kseq = KSEQ_CPU(i);
386 		if (kseq->ksq_load > high_load) {
387 			high_load = kseq->ksq_load;
388 			high_cpu = i;
389 		}
390 		if (low_load == -1 || kseq->ksq_load < low_load) {
391 			low_load = kseq->ksq_load;
392 			low_cpu = i;
393 		}
394 	}
395 
396 	kseq = KSEQ_CPU(high_cpu);
397 
398 	/*
399 	 * Nothing to do.
400 	 */
401 	if (high_load < kseq->ksq_cpus + 1)
402 		goto out;
403 
404 	high_load -= kseq->ksq_cpus;
405 
406 	if (low_load >= high_load)
407 		goto out;
408 
409 	diff = high_load - low_load;
410 	move = diff / 2;
411 	if (diff & 0x1)
412 		move++;
413 
414 	for (i = 0; i < move; i++)
415 		kseq_move(kseq, low_cpu);
416 
417 out:
418 	mtx_unlock_spin(&sched_lock);
419 	callout_reset(&kseq_lb_callout, hz, kseq_balance, NULL);
420 
421 	return;
422 }
423 
424 struct kseq *
425 kseq_load_highest(void)
426 {
427 	struct kseq *kseq;
428 	int load;
429 	int cpu;
430 	int i;
431 
432 	mtx_assert(&sched_lock, MA_OWNED);
433 	cpu = 0;
434 	load = 0;
435 
436 	for (i = 0; i < mp_maxid; i++) {
437 		if (CPU_ABSENT(i) || (i & stopped_cpus) != 0)
438 			continue;
439 		kseq = KSEQ_CPU(i);
440 		if (kseq->ksq_load > load) {
441 			load = kseq->ksq_load;
442 			cpu = i;
443 		}
444 	}
445 	kseq = KSEQ_CPU(cpu);
446 
447 	if (load > kseq->ksq_cpus)
448 		return (kseq);
449 
450 	return (NULL);
451 }
452 
453 void
454 kseq_move(struct kseq *from, int cpu)
455 {
456 	struct kse *ke;
457 
458 	ke = kseq_choose(from);
459 	runq_remove(ke->ke_runq, ke);
460 	ke->ke_state = KES_THREAD;
461 	kseq_rem(from, ke);
462 
463 	ke->ke_cpu = cpu;
464 	sched_add(ke);
465 }
466 #endif
467 
468 struct kse *
469 kseq_choose(struct kseq *kseq)
470 {
471 	struct kse *ke;
472 	struct runq *swap;
473 
474 	mtx_assert(&sched_lock, MA_OWNED);
475 	swap = NULL;
476 
477 	for (;;) {
478 		ke = runq_choose(kseq->ksq_curr);
479 		if (ke == NULL) {
480 			/*
481 			 * We already swaped once and didn't get anywhere.
482 			 */
483 			if (swap)
484 				break;
485 			swap = kseq->ksq_curr;
486 			kseq->ksq_curr = kseq->ksq_next;
487 			kseq->ksq_next = swap;
488 			continue;
489 		}
490 		/*
491 		 * If we encounter a slice of 0 the kse is in a
492 		 * TIMESHARE kse group and its nice was too far out
493 		 * of the range that receives slices.
494 		 */
495 		if (ke->ke_slice == 0) {
496 			runq_remove(ke->ke_runq, ke);
497 			sched_slice(ke);
498 			ke->ke_runq = kseq->ksq_next;
499 			runq_add(ke->ke_runq, ke);
500 			continue;
501 		}
502 		return (ke);
503 	}
504 
505 	return (runq_choose(&kseq->ksq_idle));
506 }
507 
508 static void
509 kseq_setup(struct kseq *kseq)
510 {
511 	runq_init(&kseq->ksq_timeshare[0]);
512 	runq_init(&kseq->ksq_timeshare[1]);
513 	runq_init(&kseq->ksq_idle);
514 
515 	kseq->ksq_curr = &kseq->ksq_timeshare[0];
516 	kseq->ksq_next = &kseq->ksq_timeshare[1];
517 
518 	kseq->ksq_loads[PRI_ITHD] = 0;
519 	kseq->ksq_loads[PRI_REALTIME] = 0;
520 	kseq->ksq_loads[PRI_TIMESHARE] = 0;
521 	kseq->ksq_loads[PRI_IDLE] = 0;
522 	kseq->ksq_load = 0;
523 #ifdef SMP
524 	kseq->ksq_rslices = 0;
525 #endif
526 }
527 
528 static void
529 sched_setup(void *dummy)
530 {
531 	int i;
532 
533 	slice_min = (hz/100);	/* 10ms */
534 	slice_max = (hz/7);	/* ~140ms */
535 
536 #ifdef SMP
537 	/* init kseqs */
538 	/* Create the idmap. */
539 #ifdef ULE_HTT_EXPERIMENTAL
540 	if (smp_topology == NULL) {
541 #else
542 	if (1) {
543 #endif
544 		for (i = 0; i < MAXCPU; i++) {
545 			kseq_setup(&kseq_cpu[i]);
546 			kseq_idmap[i] = &kseq_cpu[i];
547 			kseq_cpu[i].ksq_cpus = 1;
548 		}
549 	} else {
550 		int j;
551 
552 		for (i = 0; i < smp_topology->ct_count; i++) {
553 			struct cpu_group *cg;
554 
555 			cg = &smp_topology->ct_group[i];
556 			kseq_setup(&kseq_cpu[i]);
557 
558 			for (j = 0; j < MAXCPU; j++)
559 				if ((cg->cg_mask & (1 << j)) != 0)
560 					kseq_idmap[j] = &kseq_cpu[i];
561 			kseq_cpu[i].ksq_cpus = cg->cg_count;
562 		}
563 	}
564 	callout_init(&kseq_lb_callout, 1);
565 	kseq_balance(NULL);
566 #else
567 	kseq_setup(KSEQ_SELF());
568 #endif
569 	mtx_lock_spin(&sched_lock);
570 	kseq_add(KSEQ_SELF(), &kse0);
571 	mtx_unlock_spin(&sched_lock);
572 }
573 
574 /*
575  * Scale the scheduling priority according to the "interactivity" of this
576  * process.
577  */
578 static void
579 sched_priority(struct ksegrp *kg)
580 {
581 	int pri;
582 
583 	if (kg->kg_pri_class != PRI_TIMESHARE)
584 		return;
585 
586 	pri = SCHED_PRI_INTERACT(sched_interact_score(kg));
587 	pri += SCHED_PRI_BASE;
588 	pri += kg->kg_nice;
589 
590 	if (pri > PRI_MAX_TIMESHARE)
591 		pri = PRI_MAX_TIMESHARE;
592 	else if (pri < PRI_MIN_TIMESHARE)
593 		pri = PRI_MIN_TIMESHARE;
594 
595 	kg->kg_user_pri = pri;
596 
597 	return;
598 }
599 
600 /*
601  * Calculate a time slice based on the properties of the kseg and the runq
602  * that we're on.  This is only for PRI_TIMESHARE ksegrps.
603  */
604 static void
605 sched_slice(struct kse *ke)
606 {
607 	struct kseq *kseq;
608 	struct ksegrp *kg;
609 
610 	kg = ke->ke_ksegrp;
611 	kseq = KSEQ_CPU(ke->ke_cpu);
612 
613 	/*
614 	 * Rationale:
615 	 * KSEs in interactive ksegs get the minimum slice so that we
616 	 * quickly notice if it abuses its advantage.
617 	 *
618 	 * KSEs in non-interactive ksegs are assigned a slice that is
619 	 * based on the ksegs nice value relative to the least nice kseg
620 	 * on the run queue for this cpu.
621 	 *
622 	 * If the KSE is less nice than all others it gets the maximum
623 	 * slice and other KSEs will adjust their slice relative to
624 	 * this when they first expire.
625 	 *
626 	 * There is 20 point window that starts relative to the least
627 	 * nice kse on the run queue.  Slice size is determined by
628 	 * the kse distance from the last nice ksegrp.
629 	 *
630 	 * If you are outside of the window you will get no slice and
631 	 * you will be reevaluated each time you are selected on the
632 	 * run queue.
633 	 *
634 	 */
635 
636 	if (!SCHED_INTERACTIVE(kg)) {
637 		int nice;
638 
639 		nice = kg->kg_nice + (0 - kseq->ksq_nicemin);
640 		if (kseq->ksq_loads[PRI_TIMESHARE] == 0 ||
641 		    kg->kg_nice < kseq->ksq_nicemin)
642 			ke->ke_slice = SCHED_SLICE_MAX;
643 		else if (nice <= SCHED_PRI_NTHRESH)
644 			ke->ke_slice = SCHED_SLICE_NICE(nice);
645 		else
646 			ke->ke_slice = 0;
647 	} else
648 		ke->ke_slice = SCHED_SLICE_MIN;
649 
650 	CTR6(KTR_ULE,
651 	    "Sliced %p(%d) (nice: %d, nicemin: %d, load: %d, interactive: %d)",
652 	    ke, ke->ke_slice, kg->kg_nice, kseq->ksq_nicemin,
653 	    kseq->ksq_loads[PRI_TIMESHARE], SCHED_INTERACTIVE(kg));
654 
655 	/*
656 	 * Check to see if we need to scale back the slp and run time
657 	 * in the kg.  This will cause us to forget old interactivity
658 	 * while maintaining the current ratio.
659 	 */
660 	sched_interact_update(kg);
661 
662 	return;
663 }
664 
665 static void
666 sched_interact_update(struct ksegrp *kg)
667 {
668 	/* XXX Fixme, use a linear algorithm and not a while loop. */
669 	while ((kg->kg_runtime + kg->kg_slptime) >  SCHED_SLP_RUN_MAX) {
670 		kg->kg_runtime = (kg->kg_runtime / 5) * 4;
671 		kg->kg_slptime = (kg->kg_slptime / 5) * 4;
672 	}
673 }
674 
675 static int
676 sched_interact_score(struct ksegrp *kg)
677 {
678 	int div;
679 
680 	if (kg->kg_runtime > kg->kg_slptime) {
681 		div = max(1, kg->kg_runtime / SCHED_INTERACT_HALF);
682 		return (SCHED_INTERACT_HALF +
683 		    (SCHED_INTERACT_HALF - (kg->kg_slptime / div)));
684 	} if (kg->kg_slptime > kg->kg_runtime) {
685 		div = max(1, kg->kg_slptime / SCHED_INTERACT_HALF);
686 		return (kg->kg_runtime / div);
687 	}
688 
689 	/*
690 	 * This can happen if slptime and runtime are 0.
691 	 */
692 	return (0);
693 
694 }
695 
696 /*
697  * This is only somewhat accurate since given many processes of the same
698  * priority they will switch when their slices run out, which will be
699  * at most SCHED_SLICE_MAX.
700  */
701 int
702 sched_rr_interval(void)
703 {
704 	return (SCHED_SLICE_MAX);
705 }
706 
707 void
708 sched_pctcpu_update(struct kse *ke)
709 {
710 	/*
711 	 * Adjust counters and watermark for pctcpu calc.
712 	 */
713 
714 	/*
715 	 * Shift the tick count out so that the divide doesn't round away
716 	 * our results.
717 	 */
718 	ke->ke_ticks <<= 10;
719 	ke->ke_ticks = (ke->ke_ticks / (ke->ke_ltick - ke->ke_ftick)) *
720 		    SCHED_CPU_TICKS;
721 	ke->ke_ticks >>= 10;
722 	ke->ke_ltick = ticks;
723 	ke->ke_ftick = ke->ke_ltick - SCHED_CPU_TICKS;
724 }
725 
726 #ifdef SMP
727 /* XXX Should be changed to kseq_load_lowest() */
728 int
729 sched_pickcpu(void)
730 {
731 	struct kseq *kseq;
732 	int load;
733 	int cpu;
734 	int i;
735 
736 	mtx_assert(&sched_lock, MA_OWNED);
737 	if (!smp_started)
738 		return (0);
739 
740 	load = 0;
741 	cpu = 0;
742 
743 	for (i = 0; i < mp_maxid; i++) {
744 		if (CPU_ABSENT(i) || (i & stopped_cpus) != 0)
745 			continue;
746 		kseq = KSEQ_CPU(i);
747 		if (kseq->ksq_load < load) {
748 			cpu = i;
749 			load = kseq->ksq_load;
750 		}
751 	}
752 
753 	CTR1(KTR_RUNQ, "sched_pickcpu: %d", cpu);
754 	return (cpu);
755 }
756 #else
757 int
758 sched_pickcpu(void)
759 {
760 	return (0);
761 }
762 #endif
763 
764 void
765 sched_prio(struct thread *td, u_char prio)
766 {
767 	struct kse *ke;
768 	struct runq *rq;
769 
770 	mtx_assert(&sched_lock, MA_OWNED);
771 	ke = td->td_kse;
772 	td->td_priority = prio;
773 
774 	if (TD_ON_RUNQ(td)) {
775 		rq = ke->ke_runq;
776 
777 		runq_remove(rq, ke);
778 		runq_add(rq, ke);
779 	}
780 }
781 
782 void
783 sched_switchout(struct thread *td)
784 {
785 	struct kse *ke;
786 
787 	mtx_assert(&sched_lock, MA_OWNED);
788 
789 	ke = td->td_kse;
790 
791 	td->td_last_kse = ke;
792         td->td_lastcpu = td->td_oncpu;
793 	td->td_oncpu = NOCPU;
794         td->td_flags &= ~TDF_NEEDRESCHED;
795 
796 	if (TD_IS_RUNNING(td)) {
797 		/*
798 		 * This queue is always correct except for idle threads which
799 		 * have a higher priority due to priority propagation.
800 		 */
801 		if (ke->ke_ksegrp->kg_pri_class == PRI_IDLE &&
802 		    ke->ke_thread->td_priority > PRI_MIN_IDLE)
803 			ke->ke_runq = KSEQ_SELF()->ksq_curr;
804 		runq_add(ke->ke_runq, ke);
805 		/* setrunqueue(td); */
806 		return;
807 	}
808 	if (ke->ke_runq)
809 		kseq_rem(KSEQ_CPU(ke->ke_cpu), ke);
810 	/*
811 	 * We will not be on the run queue. So we must be
812 	 * sleeping or similar.
813 	 */
814 	if (td->td_proc->p_flag & P_SA)
815 		kse_reassign(ke);
816 }
817 
818 void
819 sched_switchin(struct thread *td)
820 {
821 	/* struct kse *ke = td->td_kse; */
822 	mtx_assert(&sched_lock, MA_OWNED);
823 
824 	td->td_oncpu = PCPU_GET(cpuid);
825 }
826 
827 void
828 sched_nice(struct ksegrp *kg, int nice)
829 {
830 	struct kse *ke;
831 	struct thread *td;
832 	struct kseq *kseq;
833 
834 	PROC_LOCK_ASSERT(kg->kg_proc, MA_OWNED);
835 	mtx_assert(&sched_lock, MA_OWNED);
836 	/*
837 	 * We need to adjust the nice counts for running KSEs.
838 	 */
839 	if (kg->kg_pri_class == PRI_TIMESHARE)
840 		FOREACH_KSE_IN_GROUP(kg, ke) {
841 			if (ke->ke_runq == NULL)
842 				continue;
843 			kseq = KSEQ_CPU(ke->ke_cpu);
844 			kseq_nice_rem(kseq, kg->kg_nice);
845 			kseq_nice_add(kseq, nice);
846 		}
847 	kg->kg_nice = nice;
848 	sched_priority(kg);
849 	FOREACH_THREAD_IN_GROUP(kg, td)
850 		td->td_flags |= TDF_NEEDRESCHED;
851 }
852 
853 void
854 sched_sleep(struct thread *td, u_char prio)
855 {
856 	mtx_assert(&sched_lock, MA_OWNED);
857 
858 	td->td_slptime = ticks;
859 	td->td_priority = prio;
860 
861 	CTR2(KTR_ULE, "sleep kse %p (tick: %d)",
862 	    td->td_kse, td->td_slptime);
863 }
864 
865 void
866 sched_wakeup(struct thread *td)
867 {
868 	mtx_assert(&sched_lock, MA_OWNED);
869 
870 	/*
871 	 * Let the kseg know how long we slept for.  This is because process
872 	 * interactivity behavior is modeled in the kseg.
873 	 */
874 	if (td->td_slptime) {
875 		struct ksegrp *kg;
876 		int hzticks;
877 
878 		kg = td->td_ksegrp;
879 		hzticks = ticks - td->td_slptime;
880 		kg->kg_slptime += hzticks << 10;
881 		sched_interact_update(kg);
882 		sched_priority(kg);
883 		if (td->td_kse)
884 			sched_slice(td->td_kse);
885 		CTR2(KTR_ULE, "wakeup kse %p (%d ticks)",
886 		    td->td_kse, hzticks);
887 		td->td_slptime = 0;
888 	}
889 	setrunqueue(td);
890         if (td->td_priority < curthread->td_priority)
891                 curthread->td_flags |= TDF_NEEDRESCHED;
892 }
893 
894 /*
895  * Penalize the parent for creating a new child and initialize the child's
896  * priority.
897  */
898 void
899 sched_fork(struct proc *p, struct proc *p1)
900 {
901 
902 	mtx_assert(&sched_lock, MA_OWNED);
903 
904 	sched_fork_ksegrp(FIRST_KSEGRP_IN_PROC(p), FIRST_KSEGRP_IN_PROC(p1));
905 	sched_fork_kse(FIRST_KSE_IN_PROC(p), FIRST_KSE_IN_PROC(p1));
906 	sched_fork_thread(FIRST_THREAD_IN_PROC(p), FIRST_THREAD_IN_PROC(p1));
907 }
908 
909 void
910 sched_fork_kse(struct kse *ke, struct kse *child)
911 {
912 
913 	child->ke_slice = 1;	/* Attempt to quickly learn interactivity. */
914 	child->ke_cpu = ke->ke_cpu; /* sched_pickcpu(); */
915 	child->ke_runq = NULL;
916 
917 	/*
918 	 * Claim that we've been running for one second for statistical
919 	 * purposes.
920 	 */
921 	child->ke_ticks = 0;
922 	child->ke_ltick = ticks;
923 	child->ke_ftick = ticks - hz;
924 }
925 
926 void
927 sched_fork_ksegrp(struct ksegrp *kg, struct ksegrp *child)
928 {
929 
930 	PROC_LOCK_ASSERT(child->kg_proc, MA_OWNED);
931 	/* XXX Need something better here */
932 
933 	child->kg_slptime = kg->kg_slptime / SCHED_SLP_RUN_THROTTLE;
934 	child->kg_runtime = kg->kg_runtime / SCHED_SLP_RUN_THROTTLE;
935 	kg->kg_runtime += tickincr << 10;
936 	sched_interact_update(kg);
937 
938 	child->kg_user_pri = kg->kg_user_pri;
939 	child->kg_nice = kg->kg_nice;
940 }
941 
942 void
943 sched_fork_thread(struct thread *td, struct thread *child)
944 {
945 }
946 
947 void
948 sched_class(struct ksegrp *kg, int class)
949 {
950 	struct kseq *kseq;
951 	struct kse *ke;
952 
953 	mtx_assert(&sched_lock, MA_OWNED);
954 	if (kg->kg_pri_class == class)
955 		return;
956 
957 	FOREACH_KSE_IN_GROUP(kg, ke) {
958 		if (ke->ke_state != KES_ONRUNQ &&
959 		    ke->ke_state != KES_THREAD)
960 			continue;
961 		kseq = KSEQ_CPU(ke->ke_cpu);
962 
963 		kseq->ksq_loads[PRI_BASE(kg->kg_pri_class)]--;
964 		kseq->ksq_loads[PRI_BASE(class)]++;
965 
966 		if (kg->kg_pri_class == PRI_TIMESHARE)
967 			kseq_nice_rem(kseq, kg->kg_nice);
968 		else if (class == PRI_TIMESHARE)
969 			kseq_nice_add(kseq, kg->kg_nice);
970 	}
971 
972 	kg->kg_pri_class = class;
973 }
974 
975 /*
976  * Return some of the child's priority and interactivity to the parent.
977  */
978 void
979 sched_exit(struct proc *p, struct proc *child)
980 {
981 	/* XXX Need something better here */
982 	mtx_assert(&sched_lock, MA_OWNED);
983 	sched_exit_kse(FIRST_KSE_IN_PROC(p), FIRST_KSE_IN_PROC(child));
984 	sched_exit_ksegrp(FIRST_KSEGRP_IN_PROC(p), FIRST_KSEGRP_IN_PROC(child));
985 }
986 
987 void
988 sched_exit_kse(struct kse *ke, struct kse *child)
989 {
990 	kseq_rem(KSEQ_CPU(child->ke_cpu), child);
991 }
992 
993 void
994 sched_exit_ksegrp(struct ksegrp *kg, struct ksegrp *child)
995 {
996 	/* kg->kg_slptime += child->kg_slptime; */
997 	kg->kg_runtime += child->kg_runtime;
998 	sched_interact_update(kg);
999 }
1000 
1001 void
1002 sched_exit_thread(struct thread *td, struct thread *child)
1003 {
1004 }
1005 
1006 void
1007 sched_clock(struct kse *ke)
1008 {
1009 	struct kseq *kseq;
1010 	struct ksegrp *kg;
1011 	struct thread *td;
1012 #if 0
1013 	struct kse *nke;
1014 #endif
1015 
1016 	/*
1017 	 * sched_setup() apparently happens prior to stathz being set.  We
1018 	 * need to resolve the timers earlier in the boot so we can avoid
1019 	 * calculating this here.
1020 	 */
1021 	if (realstathz == 0) {
1022 		realstathz = stathz ? stathz : hz;
1023 		tickincr = hz / realstathz;
1024 		/*
1025 		 * XXX This does not work for values of stathz that are much
1026 		 * larger than hz.
1027 		 */
1028 		if (tickincr == 0)
1029 			tickincr = 1;
1030 	}
1031 
1032 	td = ke->ke_thread;
1033 	kg = ke->ke_ksegrp;
1034 
1035 	mtx_assert(&sched_lock, MA_OWNED);
1036 	KASSERT((td != NULL), ("schedclock: null thread pointer"));
1037 
1038 	/* Adjust ticks for pctcpu */
1039 	ke->ke_ticks++;
1040 	ke->ke_ltick = ticks;
1041 
1042 	/* Go up to one second beyond our max and then trim back down */
1043 	if (ke->ke_ftick + SCHED_CPU_TICKS + hz < ke->ke_ltick)
1044 		sched_pctcpu_update(ke);
1045 
1046 	if (td->td_flags & TDF_IDLETD)
1047 		return;
1048 
1049 	CTR4(KTR_ULE, "Tick kse %p (slice: %d, slptime: %d, runtime: %d)",
1050 	    ke, ke->ke_slice, kg->kg_slptime >> 10, kg->kg_runtime >> 10);
1051 
1052 	/*
1053 	 * We only do slicing code for TIMESHARE ksegrps.
1054 	 */
1055 	if (kg->kg_pri_class != PRI_TIMESHARE)
1056 		return;
1057 	/*
1058 	 * Check for a higher priority task on the run queue.  This can happen
1059 	 * on SMP if another processor woke up a process on our runq.
1060 	 */
1061 	kseq = KSEQ_SELF();
1062 #if 0
1063 	if (kseq->ksq_load > 1 && (nke = kseq_choose(kseq)) != NULL) {
1064 		if (sched_strict &&
1065 		    nke->ke_thread->td_priority < td->td_priority)
1066 			td->td_flags |= TDF_NEEDRESCHED;
1067 		else if (nke->ke_thread->td_priority <
1068 		    td->td_priority SCHED_PRIO_SLOP)
1069 
1070 		if (nke->ke_thread->td_priority < td->td_priority)
1071 			td->td_flags |= TDF_NEEDRESCHED;
1072 	}
1073 #endif
1074 	/*
1075 	 * We used a tick charge it to the ksegrp so that we can compute our
1076 	 * interactivity.
1077 	 */
1078 	kg->kg_runtime += tickincr << 10;
1079 	sched_interact_update(kg);
1080 
1081 	/*
1082 	 * We used up one time slice.
1083 	 */
1084 	ke->ke_slice--;
1085 #ifdef SMP
1086 	kseq->ksq_rslices--;
1087 #endif
1088 
1089 	if (ke->ke_slice > 0)
1090 		return;
1091 	/*
1092 	 * We're out of time, recompute priorities and requeue.
1093 	 */
1094 	kseq_rem(kseq, ke);
1095 	sched_priority(kg);
1096 	sched_slice(ke);
1097 	if (SCHED_CURR(kg, ke))
1098 		ke->ke_runq = kseq->ksq_curr;
1099 	else
1100 		ke->ke_runq = kseq->ksq_next;
1101 	kseq_add(kseq, ke);
1102 	td->td_flags |= TDF_NEEDRESCHED;
1103 }
1104 
1105 int
1106 sched_runnable(void)
1107 {
1108 	struct kseq *kseq;
1109 	int load;
1110 
1111 	load = 1;
1112 
1113 	mtx_lock_spin(&sched_lock);
1114 	kseq = KSEQ_SELF();
1115 
1116 	if (kseq->ksq_load)
1117 		goto out;
1118 #ifdef SMP
1119 	/*
1120 	 * For SMP we may steal other processor's KSEs.  Just search until we
1121 	 * verify that at least on other cpu has a runnable task.
1122 	 */
1123 	if (smp_started) {
1124 		int i;
1125 
1126 		for (i = 0; i < mp_maxid; i++) {
1127 			if (CPU_ABSENT(i) || (i & stopped_cpus) != 0)
1128 				continue;
1129 			kseq = KSEQ_CPU(i);
1130 			if (kseq->ksq_load > kseq->ksq_cpus)
1131 				goto out;
1132 		}
1133 	}
1134 #endif
1135 	load = 0;
1136 out:
1137 	mtx_unlock_spin(&sched_lock);
1138 	return (load);
1139 }
1140 
1141 void
1142 sched_userret(struct thread *td)
1143 {
1144 	struct ksegrp *kg;
1145 	struct kseq *kseq;
1146 	struct kse *ke;
1147 
1148 	kg = td->td_ksegrp;
1149 
1150 	if (td->td_priority != kg->kg_user_pri) {
1151 		mtx_lock_spin(&sched_lock);
1152 		td->td_priority = kg->kg_user_pri;
1153 		kseq = KSEQ_SELF();
1154 		if (td->td_ksegrp->kg_pri_class == PRI_TIMESHARE &&
1155 #ifdef SMP
1156 		    kseq->ksq_load > kseq->ksq_cpus &&
1157 #else
1158 		    kseq->ksq_load > 1 &&
1159 #endif
1160 		    (ke = kseq_choose(kseq)) != NULL &&
1161 		    ke->ke_thread->td_priority < td->td_priority)
1162 			curthread->td_flags |= TDF_NEEDRESCHED;
1163 		mtx_unlock_spin(&sched_lock);
1164 	}
1165 }
1166 
1167 struct kse *
1168 sched_choose(void)
1169 {
1170 	struct kseq *kseq;
1171 	struct kse *ke;
1172 
1173 	mtx_assert(&sched_lock, MA_OWNED);
1174 #ifdef SMP
1175 retry:
1176 #endif
1177 	kseq = KSEQ_SELF();
1178 	ke = kseq_choose(kseq);
1179 	if (ke) {
1180 		runq_remove(ke->ke_runq, ke);
1181 		ke->ke_state = KES_THREAD;
1182 
1183 		if (ke->ke_ksegrp->kg_pri_class == PRI_TIMESHARE) {
1184 			CTR4(KTR_ULE, "Run kse %p from %p (slice: %d, pri: %d)",
1185 			    ke, ke->ke_runq, ke->ke_slice,
1186 			    ke->ke_thread->td_priority);
1187 		}
1188 		return (ke);
1189 	}
1190 
1191 #ifdef SMP
1192 	if (smp_started) {
1193 		/*
1194 		 * Find the cpu with the highest load and steal one proc.
1195 		 */
1196 		if ((kseq = kseq_load_highest()) == NULL)
1197 			return (NULL);
1198 
1199 		/*
1200 		 * Remove this kse from this kseq and runq and then requeue
1201 		 * on the current processor.  Then we will dequeue it
1202 		 * normally above.
1203 		 */
1204 		kseq_move(kseq, PCPU_GET(cpuid));
1205 		goto retry;
1206 	}
1207 #endif
1208 
1209 	return (NULL);
1210 }
1211 
1212 void
1213 sched_add(struct kse *ke)
1214 {
1215 	struct kseq *kseq;
1216 	struct ksegrp *kg;
1217 
1218 	mtx_assert(&sched_lock, MA_OWNED);
1219 	KASSERT((ke->ke_thread != NULL), ("sched_add: No thread on KSE"));
1220 	KASSERT((ke->ke_thread->td_kse != NULL),
1221 	    ("sched_add: No KSE on thread"));
1222 	KASSERT(ke->ke_state != KES_ONRUNQ,
1223 	    ("sched_add: kse %p (%s) already in run queue", ke,
1224 	    ke->ke_proc->p_comm));
1225 	KASSERT(ke->ke_proc->p_sflag & PS_INMEM,
1226 	    ("sched_add: process swapped out"));
1227 	KASSERT(ke->ke_runq == NULL,
1228 	    ("sched_add: KSE %p is still assigned to a run queue", ke));
1229 
1230 	kg = ke->ke_ksegrp;
1231 
1232 	switch (PRI_BASE(kg->kg_pri_class)) {
1233 	case PRI_ITHD:
1234 	case PRI_REALTIME:
1235 		kseq = KSEQ_SELF();
1236 		ke->ke_runq = kseq->ksq_curr;
1237 		ke->ke_slice = SCHED_SLICE_MAX;
1238 		ke->ke_cpu = PCPU_GET(cpuid);
1239 		break;
1240 	case PRI_TIMESHARE:
1241 		kseq = KSEQ_CPU(ke->ke_cpu);
1242 		if (SCHED_CURR(kg, ke))
1243 			ke->ke_runq = kseq->ksq_curr;
1244 		else
1245 			ke->ke_runq = kseq->ksq_next;
1246 		break;
1247 	case PRI_IDLE:
1248 		kseq = KSEQ_CPU(ke->ke_cpu);
1249 		/*
1250 		 * This is for priority prop.
1251 		 */
1252 		if (ke->ke_thread->td_priority > PRI_MIN_IDLE)
1253 			ke->ke_runq = kseq->ksq_curr;
1254 		else
1255 			ke->ke_runq = &kseq->ksq_idle;
1256 		ke->ke_slice = SCHED_SLICE_MIN;
1257 		break;
1258 	default:
1259 		panic("Unknown pri class.\n");
1260 		break;
1261 	}
1262 
1263 	ke->ke_ksegrp->kg_runq_kses++;
1264 	ke->ke_state = KES_ONRUNQ;
1265 
1266 	runq_add(ke->ke_runq, ke);
1267 	kseq_add(kseq, ke);
1268 }
1269 
1270 void
1271 sched_rem(struct kse *ke)
1272 {
1273 	struct kseq *kseq;
1274 
1275 	mtx_assert(&sched_lock, MA_OWNED);
1276 	KASSERT((ke->ke_state == KES_ONRUNQ), ("KSE not on run queue"));
1277 
1278 	ke->ke_state = KES_THREAD;
1279 	ke->ke_ksegrp->kg_runq_kses--;
1280 	kseq = KSEQ_CPU(ke->ke_cpu);
1281 	runq_remove(ke->ke_runq, ke);
1282 	kseq_rem(kseq, ke);
1283 }
1284 
1285 fixpt_t
1286 sched_pctcpu(struct kse *ke)
1287 {
1288 	fixpt_t pctcpu;
1289 
1290 	pctcpu = 0;
1291 
1292 	mtx_lock_spin(&sched_lock);
1293 	if (ke->ke_ticks) {
1294 		int rtick;
1295 
1296 		/*
1297 		 * Don't update more frequently than twice a second.  Allowing
1298 		 * this causes the cpu usage to decay away too quickly due to
1299 		 * rounding errors.
1300 		 */
1301 		if (ke->ke_ltick < (ticks - (hz / 2)))
1302 			sched_pctcpu_update(ke);
1303 
1304 		/* How many rtick per second ? */
1305 		rtick = min(ke->ke_ticks / SCHED_CPU_TIME, SCHED_CPU_TICKS);
1306 		pctcpu = (FSCALE * ((FSCALE * rtick)/realstathz)) >> FSHIFT;
1307 	}
1308 
1309 	ke->ke_proc->p_swtime = ke->ke_ltick - ke->ke_ftick;
1310 	mtx_unlock_spin(&sched_lock);
1311 
1312 	return (pctcpu);
1313 }
1314 
1315 int
1316 sched_sizeof_kse(void)
1317 {
1318 	return (sizeof(struct kse) + sizeof(struct ke_sched));
1319 }
1320 
1321 int
1322 sched_sizeof_ksegrp(void)
1323 {
1324 	return (sizeof(struct ksegrp) + sizeof(struct kg_sched));
1325 }
1326 
1327 int
1328 sched_sizeof_proc(void)
1329 {
1330 	return (sizeof(struct proc));
1331 }
1332 
1333 int
1334 sched_sizeof_thread(void)
1335 {
1336 	return (sizeof(struct thread) + sizeof(struct td_sched));
1337 }
1338