xref: /freebsd/sys/kern/sched_ule.c (revision b90816f188792ddbdd0b7537f81593ef694ba39d)
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  * $FreeBSD$
27  */
28 
29 #include <sys/param.h>
30 #include <sys/systm.h>
31 #include <sys/kernel.h>
32 #include <sys/ktr.h>
33 #include <sys/lock.h>
34 #include <sys/mutex.h>
35 #include <sys/proc.h>
36 #include <sys/resource.h>
37 #include <sys/sched.h>
38 #include <sys/smp.h>
39 #include <sys/sx.h>
40 #include <sys/sysctl.h>
41 #include <sys/sysproto.h>
42 #include <sys/vmmeter.h>
43 #ifdef DDB
44 #include <ddb/ddb.h>
45 #endif
46 #ifdef KTRACE
47 #include <sys/uio.h>
48 #include <sys/ktrace.h>
49 #endif
50 
51 #include <machine/cpu.h>
52 
53 #define KTR_ULE         KTR_NFS
54 
55 /* decay 95% of `p_pctcpu' in 60 seconds; see CCPU_SHIFT before changing */
56 /* XXX This is bogus compatability crap for ps */
57 static fixpt_t  ccpu = 0.95122942450071400909 * FSCALE; /* exp(-1/20) */
58 SYSCTL_INT(_kern, OID_AUTO, ccpu, CTLFLAG_RD, &ccpu, 0, "");
59 
60 static void sched_setup(void *dummy);
61 SYSINIT(sched_setup, SI_SUB_RUN_QUEUE, SI_ORDER_FIRST, sched_setup, NULL)
62 
63 static SYSCTL_NODE(_kern, OID_AUTO, sched, CTLFLAG_RW, 0, "SCHED");
64 
65 static int sched_strict;
66 SYSCTL_INT(_kern_sched, OID_AUTO, strict, CTLFLAG_RD, &sched_strict, 0, "");
67 
68 static int slice_min = 1;
69 SYSCTL_INT(_kern_sched, OID_AUTO, slice_min, CTLFLAG_RW, &slice_min, 0, "");
70 
71 static int slice_max = 2;
72 SYSCTL_INT(_kern_sched, OID_AUTO, slice_max, CTLFLAG_RW, &slice_max, 0, "");
73 
74 int realstathz;
75 int tickincr = 1;
76 
77 /*
78  * These datastructures are allocated within their parent datastructure but
79  * are scheduler specific.
80  */
81 
82 struct ke_sched {
83 	int		ske_slice;
84 	struct runq	*ske_runq;
85 	/* The following variables are only used for pctcpu calculation */
86 	int		ske_ltick;	/* Last tick that we were running on */
87 	int		ske_ftick;	/* First tick that we were running on */
88 	int		ske_ticks;	/* Tick count */
89 	/* CPU that we have affinity for. */
90 	u_char		ske_cpu;
91 };
92 #define	ke_slice	ke_sched->ske_slice
93 #define	ke_runq		ke_sched->ske_runq
94 #define	ke_ltick	ke_sched->ske_ltick
95 #define	ke_ftick	ke_sched->ske_ftick
96 #define	ke_ticks	ke_sched->ske_ticks
97 #define	ke_cpu		ke_sched->ske_cpu
98 
99 struct kg_sched {
100 	int	skg_slptime;		/* Number of ticks we vol. slept */
101 	int	skg_runtime;		/* Number of ticks we were running */
102 };
103 #define	kg_slptime	kg_sched->skg_slptime
104 #define	kg_runtime	kg_sched->skg_runtime
105 
106 struct td_sched {
107 	int	std_slptime;
108 };
109 #define	td_slptime	td_sched->std_slptime
110 
111 struct td_sched td_sched;
112 struct ke_sched ke_sched;
113 struct kg_sched kg_sched;
114 
115 struct ke_sched *kse0_sched = &ke_sched;
116 struct kg_sched *ksegrp0_sched = &kg_sched;
117 struct p_sched *proc0_sched = NULL;
118 struct td_sched *thread0_sched = &td_sched;
119 
120 /*
121  * This priority range has 20 priorities on either end that are reachable
122  * only through nice values.
123  *
124  * PRI_RANGE:	Total priority range for timeshare threads.
125  * PRI_NRESV:	Reserved priorities for nice.
126  * PRI_BASE:	The start of the dynamic range.
127  * DYN_RANGE:	Number of priorities that are available int the dynamic
128  *		priority range.
129  */
130 #define	SCHED_PRI_RANGE		(PRI_MAX_TIMESHARE - PRI_MIN_TIMESHARE + 1)
131 #define	SCHED_PRI_NRESV		PRIO_TOTAL
132 #define	SCHED_PRI_NHALF		(PRIO_TOTAL / 2)
133 #define	SCHED_PRI_NTHRESH	(SCHED_PRI_NHALF - 1)
134 #define	SCHED_PRI_BASE		((SCHED_PRI_NRESV / 2) + PRI_MIN_TIMESHARE)
135 #define	SCHED_DYN_RANGE		(SCHED_PRI_RANGE - SCHED_PRI_NRESV)
136 #define	SCHED_PRI_INTERACT(score)					\
137     ((score) * SCHED_DYN_RANGE / SCHED_INTERACT_RANGE)
138 
139 /*
140  * These determine the interactivity of a process.
141  *
142  * SLP_RUN_MAX:	Maximum amount of sleep time + run time we'll accumulate
143  *		before throttling back.
144  * SLP_RUN_THROTTLE:	Divisor for reducing slp/run time.
145  * INTERACT_RANGE:	Range of interactivity values.  Smaller is better.
146  * INTERACT_HALF:	Convenience define, half of the interactivity range.
147  * INTERACT_THRESH:	Threshhold for placement on the current runq.
148  */
149 #define	SCHED_SLP_RUN_MAX	((hz / 10) << 10)
150 #define	SCHED_SLP_RUN_THROTTLE	(10)
151 #define	SCHED_INTERACT_RANGE	(100)
152 #define	SCHED_INTERACT_HALF	(SCHED_INTERACT_RANGE / 2)
153 #define	SCHED_INTERACT_THRESH	(10)
154 
155 /*
156  * These parameters and macros determine the size of the time slice that is
157  * granted to each thread.
158  *
159  * SLICE_MIN:	Minimum time slice granted, in units of ticks.
160  * SLICE_MAX:	Maximum time slice granted.
161  * SLICE_RANGE:	Range of available time slices scaled by hz.
162  * SLICE_SCALE:	The number slices granted per val in the range of [0, max].
163  * SLICE_NICE:  Determine the amount of slice granted to a scaled nice.
164  */
165 #define	SCHED_SLICE_MIN			(slice_min)
166 #define	SCHED_SLICE_MAX			(slice_max)
167 #define	SCHED_SLICE_RANGE		(SCHED_SLICE_MAX - SCHED_SLICE_MIN + 1)
168 #define	SCHED_SLICE_SCALE(val, max)	(((val) * SCHED_SLICE_RANGE) / (max))
169 #define	SCHED_SLICE_NICE(nice)						\
170     (SCHED_SLICE_MAX - SCHED_SLICE_SCALE((nice), SCHED_PRI_NTHRESH))
171 
172 /*
173  * This macro determines whether or not the kse belongs on the current or
174  * next run queue.
175  *
176  * XXX nice value should effect how interactive a kg is.
177  */
178 #define	SCHED_INTERACTIVE(kg)						\
179     (sched_interact_score(kg) < SCHED_INTERACT_THRESH)
180 #define	SCHED_CURR(kg, ke)						\
181     (ke->ke_thread->td_priority < PRI_MIN_TIMESHARE || SCHED_INTERACTIVE(kg))
182 
183 /*
184  * Cpu percentage computation macros and defines.
185  *
186  * SCHED_CPU_TIME:	Number of seconds to average the cpu usage across.
187  * SCHED_CPU_TICKS:	Number of hz ticks to average the cpu usage across.
188  */
189 
190 #define	SCHED_CPU_TIME	10
191 #define	SCHED_CPU_TICKS	(hz * SCHED_CPU_TIME)
192 
193 /*
194  * kseq - per processor runqs and statistics.
195  */
196 
197 #define	KSEQ_NCLASS	(PRI_IDLE + 1)	/* Number of run classes. */
198 
199 struct kseq {
200 	struct runq	ksq_idle;		/* Queue of IDLE threads. */
201 	struct runq	ksq_timeshare[2];	/* Run queues for !IDLE. */
202 	struct runq	*ksq_next;		/* Next timeshare queue. */
203 	struct runq	*ksq_curr;		/* Current queue. */
204 	int		ksq_loads[KSEQ_NCLASS];	/* Load for each class */
205 	int		ksq_load;		/* Aggregate load. */
206 	short		ksq_nice[PRIO_TOTAL + 1]; /* KSEs in each nice bin. */
207 	short		ksq_nicemin;		/* Least nice. */
208 #ifdef SMP
209 	unsigned int	ksq_rslices;	/* Slices on run queue */
210 #endif
211 };
212 
213 /*
214  * One kse queue per processor.
215  */
216 #ifdef SMP
217 struct kseq	kseq_cpu[MAXCPU];
218 #define	KSEQ_SELF()	(&kseq_cpu[PCPU_GET(cpuid)])
219 #define	KSEQ_CPU(x)	(&kseq_cpu[(x)])
220 #else
221 struct kseq	kseq_cpu;
222 #define	KSEQ_SELF()	(&kseq_cpu)
223 #define	KSEQ_CPU(x)	(&kseq_cpu)
224 #endif
225 
226 static void sched_slice(struct kse *ke);
227 static void sched_priority(struct ksegrp *kg);
228 static int sched_interact_score(struct ksegrp *kg);
229 void sched_pctcpu_update(struct kse *ke);
230 int sched_pickcpu(void);
231 
232 /* Operations on per processor queues */
233 static struct kse * kseq_choose(struct kseq *kseq);
234 static void kseq_setup(struct kseq *kseq);
235 static void kseq_add(struct kseq *kseq, struct kse *ke);
236 static void kseq_rem(struct kseq *kseq, struct kse *ke);
237 static void kseq_nice_add(struct kseq *kseq, int nice);
238 static void kseq_nice_rem(struct kseq *kseq, int nice);
239 void kseq_print(int cpu);
240 #ifdef SMP
241 struct kseq * kseq_load_highest(void);
242 #endif
243 
244 void
245 kseq_print(int cpu)
246 {
247 	struct kseq *kseq;
248 	int i;
249 
250 	kseq = KSEQ_CPU(cpu);
251 
252 	printf("kseq:\n");
253 	printf("\tload:           %d\n", kseq->ksq_load);
254 	printf("\tload ITHD:      %d\n", kseq->ksq_loads[PRI_ITHD]);
255 	printf("\tload REALTIME:  %d\n", kseq->ksq_loads[PRI_REALTIME]);
256 	printf("\tload TIMESHARE: %d\n", kseq->ksq_loads[PRI_TIMESHARE]);
257 	printf("\tload IDLE:      %d\n", kseq->ksq_loads[PRI_IDLE]);
258 	printf("\tnicemin:\t%d\n", kseq->ksq_nicemin);
259 	printf("\tnice counts:\n");
260 	for (i = 0; i < PRIO_TOTAL + 1; i++)
261 		if (kseq->ksq_nice[i])
262 			printf("\t\t%d = %d\n",
263 			    i - SCHED_PRI_NHALF, kseq->ksq_nice[i]);
264 }
265 
266 static void
267 kseq_add(struct kseq *kseq, struct kse *ke)
268 {
269 	mtx_assert(&sched_lock, MA_OWNED);
270 	kseq->ksq_loads[PRI_BASE(ke->ke_ksegrp->kg_pri_class)]++;
271 	kseq->ksq_load++;
272 	if (ke->ke_ksegrp->kg_pri_class == PRI_TIMESHARE)
273 	CTR6(KTR_ULE, "Add kse %p to %p (slice: %d, pri: %d, nice: %d(%d))",
274 	    ke, ke->ke_runq, ke->ke_slice, ke->ke_thread->td_priority,
275 	    ke->ke_ksegrp->kg_nice, kseq->ksq_nicemin);
276 	if (ke->ke_ksegrp->kg_pri_class == PRI_TIMESHARE)
277 		kseq_nice_add(kseq, ke->ke_ksegrp->kg_nice);
278 #ifdef SMP
279 	kseq->ksq_rslices += ke->ke_slice;
280 #endif
281 }
282 
283 static void
284 kseq_rem(struct kseq *kseq, struct kse *ke)
285 {
286 	mtx_assert(&sched_lock, MA_OWNED);
287 	kseq->ksq_loads[PRI_BASE(ke->ke_ksegrp->kg_pri_class)]--;
288 	kseq->ksq_load--;
289 	ke->ke_runq = NULL;
290 	if (ke->ke_ksegrp->kg_pri_class == PRI_TIMESHARE)
291 		kseq_nice_rem(kseq, ke->ke_ksegrp->kg_nice);
292 #ifdef SMP
293 	kseq->ksq_rslices -= ke->ke_slice;
294 #endif
295 }
296 
297 static void
298 kseq_nice_add(struct kseq *kseq, int nice)
299 {
300 	mtx_assert(&sched_lock, MA_OWNED);
301 	/* Normalize to zero. */
302 	kseq->ksq_nice[nice + SCHED_PRI_NHALF]++;
303 	if (nice < kseq->ksq_nicemin || kseq->ksq_loads[PRI_TIMESHARE] == 1)
304 		kseq->ksq_nicemin = nice;
305 }
306 
307 static void
308 kseq_nice_rem(struct kseq *kseq, int nice)
309 {
310 	int n;
311 
312 	mtx_assert(&sched_lock, MA_OWNED);
313 	/* Normalize to zero. */
314 	n = nice + SCHED_PRI_NHALF;
315 	kseq->ksq_nice[n]--;
316 	KASSERT(kseq->ksq_nice[n] >= 0, ("Negative nice count."));
317 
318 	/*
319 	 * If this wasn't the smallest nice value or there are more in
320 	 * this bucket we can just return.  Otherwise we have to recalculate
321 	 * the smallest nice.
322 	 */
323 	if (nice != kseq->ksq_nicemin ||
324 	    kseq->ksq_nice[n] != 0 ||
325 	    kseq->ksq_loads[PRI_TIMESHARE] == 0)
326 		return;
327 
328 	for (; n < SCHED_PRI_NRESV + 1; n++)
329 		if (kseq->ksq_nice[n]) {
330 			kseq->ksq_nicemin = n - SCHED_PRI_NHALF;
331 			return;
332 		}
333 }
334 
335 #ifdef SMP
336 struct kseq *
337 kseq_load_highest(void)
338 {
339 	struct kseq *kseq;
340 	int load;
341 	int cpu;
342 	int i;
343 
344 	mtx_assert(&sched_lock, MA_OWNED);
345 	cpu = 0;
346 	load = 0;
347 
348 	for (i = 0; i < mp_maxid; i++) {
349 		if (CPU_ABSENT(i))
350 			continue;
351 		kseq = KSEQ_CPU(i);
352 		if (kseq->ksq_load > load) {
353 			load = kseq->ksq_load;
354 			cpu = i;
355 		}
356 	}
357 	if (load > 1)
358 		return (KSEQ_CPU(cpu));
359 
360 	return (NULL);
361 }
362 #endif
363 
364 struct kse *
365 kseq_choose(struct kseq *kseq)
366 {
367 	struct kse *ke;
368 	struct runq *swap;
369 
370 	mtx_assert(&sched_lock, MA_OWNED);
371 	swap = NULL;
372 
373 	for (;;) {
374 		ke = runq_choose(kseq->ksq_curr);
375 		if (ke == NULL) {
376 			/*
377 			 * We already swaped once and didn't get anywhere.
378 			 */
379 			if (swap)
380 				break;
381 			swap = kseq->ksq_curr;
382 			kseq->ksq_curr = kseq->ksq_next;
383 			kseq->ksq_next = swap;
384 			continue;
385 		}
386 		/*
387 		 * If we encounter a slice of 0 the kse is in a
388 		 * TIMESHARE kse group and its nice was too far out
389 		 * of the range that receives slices.
390 		 */
391 		if (ke->ke_slice == 0) {
392 			runq_remove(ke->ke_runq, ke);
393 			sched_slice(ke);
394 			ke->ke_runq = kseq->ksq_next;
395 			runq_add(ke->ke_runq, ke);
396 			continue;
397 		}
398 		return (ke);
399 	}
400 
401 	return (runq_choose(&kseq->ksq_idle));
402 }
403 
404 static void
405 kseq_setup(struct kseq *kseq)
406 {
407 	runq_init(&kseq->ksq_timeshare[0]);
408 	runq_init(&kseq->ksq_timeshare[1]);
409 	runq_init(&kseq->ksq_idle);
410 
411 	kseq->ksq_curr = &kseq->ksq_timeshare[0];
412 	kseq->ksq_next = &kseq->ksq_timeshare[1];
413 
414 	kseq->ksq_loads[PRI_ITHD] = 0;
415 	kseq->ksq_loads[PRI_REALTIME] = 0;
416 	kseq->ksq_loads[PRI_TIMESHARE] = 0;
417 	kseq->ksq_loads[PRI_IDLE] = 0;
418 	kseq->ksq_load = 0;
419 #ifdef SMP
420 	kseq->ksq_rslices = 0;
421 #endif
422 }
423 
424 static void
425 sched_setup(void *dummy)
426 {
427 	int i;
428 
429 	slice_min = (hz/100);
430 	slice_max = (hz/10);
431 
432 	mtx_lock_spin(&sched_lock);
433 	/* init kseqs */
434 	for (i = 0; i < MAXCPU; i++)
435 		kseq_setup(KSEQ_CPU(i));
436 
437 	kseq_add(KSEQ_SELF(), &kse0);
438 	mtx_unlock_spin(&sched_lock);
439 }
440 
441 /*
442  * Scale the scheduling priority according to the "interactivity" of this
443  * process.
444  */
445 static void
446 sched_priority(struct ksegrp *kg)
447 {
448 	int pri;
449 
450 	if (kg->kg_pri_class != PRI_TIMESHARE)
451 		return;
452 
453 	pri = SCHED_PRI_INTERACT(sched_interact_score(kg));
454 	pri += SCHED_PRI_BASE;
455 	pri += kg->kg_nice;
456 
457 	if (pri > PRI_MAX_TIMESHARE)
458 		pri = PRI_MAX_TIMESHARE;
459 	else if (pri < PRI_MIN_TIMESHARE)
460 		pri = PRI_MIN_TIMESHARE;
461 
462 	kg->kg_user_pri = pri;
463 
464 	return;
465 }
466 
467 /*
468  * Calculate a time slice based on the properties of the kseg and the runq
469  * that we're on.  This is only for PRI_TIMESHARE ksegrps.
470  */
471 static void
472 sched_slice(struct kse *ke)
473 {
474 	struct kseq *kseq;
475 	struct ksegrp *kg;
476 
477 	kg = ke->ke_ksegrp;
478 	kseq = KSEQ_CPU(ke->ke_cpu);
479 
480 	/*
481 	 * Rationale:
482 	 * KSEs in interactive ksegs get the minimum slice so that we
483 	 * quickly notice if it abuses its advantage.
484 	 *
485 	 * KSEs in non-interactive ksegs are assigned a slice that is
486 	 * based on the ksegs nice value relative to the least nice kseg
487 	 * on the run queue for this cpu.
488 	 *
489 	 * If the KSE is less nice than all others it gets the maximum
490 	 * slice and other KSEs will adjust their slice relative to
491 	 * this when they first expire.
492 	 *
493 	 * There is 20 point window that starts relative to the least
494 	 * nice kse on the run queue.  Slice size is determined by
495 	 * the kse distance from the last nice ksegrp.
496 	 *
497 	 * If you are outside of the window you will get no slice and
498 	 * you will be reevaluated each time you are selected on the
499 	 * run queue.
500 	 *
501 	 */
502 
503 	if (!SCHED_INTERACTIVE(kg)) {
504 		int nice;
505 
506 		nice = kg->kg_nice + (0 - kseq->ksq_nicemin);
507 		if (kseq->ksq_loads[PRI_TIMESHARE] == 0 ||
508 		    kg->kg_nice < kseq->ksq_nicemin)
509 			ke->ke_slice = SCHED_SLICE_MAX;
510 		else if (nice <= SCHED_PRI_NTHRESH)
511 			ke->ke_slice = SCHED_SLICE_NICE(nice);
512 		else
513 			ke->ke_slice = 0;
514 	} else
515 		ke->ke_slice = SCHED_SLICE_MIN;
516 
517 	CTR6(KTR_ULE,
518 	    "Sliced %p(%d) (nice: %d, nicemin: %d, load: %d, interactive: %d)",
519 	    ke, ke->ke_slice, kg->kg_nice, kseq->ksq_nicemin,
520 	    kseq->ksq_loads[PRI_TIMESHARE], SCHED_INTERACTIVE(kg));
521 
522 	/*
523 	 * Check to see if we need to scale back the slp and run time
524 	 * in the kg.  This will cause us to forget old interactivity
525 	 * while maintaining the current ratio.
526 	 */
527 	CTR4(KTR_ULE, "Slp vs Run %p (Slp %d, Run %d, Score %d)",
528 	    ke, kg->kg_slptime >> 10, kg->kg_runtime >> 10,
529 	    sched_interact_score(kg));
530 
531 	if ((kg->kg_runtime + kg->kg_slptime) >  SCHED_SLP_RUN_MAX) {
532 		kg->kg_runtime /= SCHED_SLP_RUN_THROTTLE;
533 		kg->kg_slptime /= SCHED_SLP_RUN_THROTTLE;
534 	}
535 	CTR4(KTR_ULE, "Slp vs Run(2) %p (Slp %d, Run %d, Score %d)",
536 	    ke, kg->kg_slptime >> 10, kg->kg_runtime >> 10,
537 	    sched_interact_score(kg));
538 
539 	return;
540 }
541 
542 static int
543 sched_interact_score(struct ksegrp *kg)
544 {
545 	int big;
546 	int small;
547 	int base;
548 
549 	if (kg->kg_runtime > kg->kg_slptime) {
550 		big = kg->kg_runtime;
551 		small = kg->kg_slptime;
552 		base = SCHED_INTERACT_HALF;
553 	} else {
554 		big = kg->kg_slptime;
555 		small = kg->kg_runtime;
556 		base = 0;
557 	}
558 
559 	big /= SCHED_INTERACT_HALF;
560 	if (big != 0)
561 		small /= big;
562 	else
563 		small = 0;
564 
565 	small += base;
566 	/* XXX Factor in nice */
567 	return (small);
568 }
569 
570 /*
571  * This is only somewhat accurate since given many processes of the same
572  * priority they will switch when their slices run out, which will be
573  * at most SCHED_SLICE_MAX.
574  */
575 int
576 sched_rr_interval(void)
577 {
578 	return (SCHED_SLICE_MAX);
579 }
580 
581 void
582 sched_pctcpu_update(struct kse *ke)
583 {
584 	/*
585 	 * Adjust counters and watermark for pctcpu calc.
586 	 *
587 	 * Shift the tick count out so that the divide doesn't round away
588 	 * our results.
589 	 */
590 	ke->ke_ticks <<= 10;
591 	ke->ke_ticks = (ke->ke_ticks / (ke->ke_ltick - ke->ke_ftick)) *
592 		    SCHED_CPU_TICKS;
593 	ke->ke_ticks >>= 10;
594 	ke->ke_ltick = ticks;
595 	ke->ke_ftick = ke->ke_ltick - SCHED_CPU_TICKS;
596 }
597 
598 #ifdef SMP
599 /* XXX Should be changed to kseq_load_lowest() */
600 int
601 sched_pickcpu(void)
602 {
603 	struct kseq *kseq;
604 	int load;
605 	int cpu;
606 	int i;
607 
608 	mtx_assert(&sched_lock, MA_OWNED);
609 	if (!smp_started)
610 		return (0);
611 
612 	load = 0;
613 	cpu = 0;
614 
615 	for (i = 0; i < mp_maxid; i++) {
616 		if (CPU_ABSENT(i))
617 			continue;
618 		kseq = KSEQ_CPU(i);
619 		if (kseq->ksq_load < load) {
620 			cpu = i;
621 			load = kseq->ksq_load;
622 		}
623 	}
624 
625 	CTR1(KTR_RUNQ, "sched_pickcpu: %d", cpu);
626 	return (cpu);
627 }
628 #else
629 int
630 sched_pickcpu(void)
631 {
632 	return (0);
633 }
634 #endif
635 
636 void
637 sched_prio(struct thread *td, u_char prio)
638 {
639 	struct kse *ke;
640 	struct runq *rq;
641 
642 	mtx_assert(&sched_lock, MA_OWNED);
643 	ke = td->td_kse;
644 	td->td_priority = prio;
645 
646 	if (TD_ON_RUNQ(td)) {
647 		rq = ke->ke_runq;
648 
649 		runq_remove(rq, ke);
650 		runq_add(rq, ke);
651 	}
652 }
653 
654 void
655 sched_switchout(struct thread *td)
656 {
657 	struct kse *ke;
658 
659 	mtx_assert(&sched_lock, MA_OWNED);
660 
661 	ke = td->td_kse;
662 
663 	td->td_last_kse = ke;
664         td->td_lastcpu = td->td_oncpu;
665 	td->td_oncpu = NOCPU;
666         td->td_flags &= ~TDF_NEEDRESCHED;
667 
668 	if (TD_IS_RUNNING(td)) {
669 		runq_add(ke->ke_runq, ke);
670 		/* setrunqueue(td); */
671 		return;
672 	}
673 	if (ke->ke_runq)
674 		kseq_rem(KSEQ_CPU(ke->ke_cpu), ke);
675 	/*
676 	 * We will not be on the run queue. So we must be
677 	 * sleeping or similar.
678 	 */
679 	if (td->td_proc->p_flag & P_THREADED)
680 		kse_reassign(ke);
681 }
682 
683 void
684 sched_switchin(struct thread *td)
685 {
686 	/* struct kse *ke = td->td_kse; */
687 	mtx_assert(&sched_lock, MA_OWNED);
688 
689 	td->td_oncpu = PCPU_GET(cpuid);
690 
691 	if (td->td_ksegrp->kg_pri_class == PRI_TIMESHARE &&
692 	    td->td_priority != td->td_ksegrp->kg_user_pri)
693 		curthread->td_flags |= TDF_NEEDRESCHED;
694 }
695 
696 void
697 sched_nice(struct ksegrp *kg, int nice)
698 {
699 	struct kse *ke;
700 	struct thread *td;
701 	struct kseq *kseq;
702 
703 	PROC_LOCK_ASSERT(kg->kg_proc, MA_OWNED);
704 	mtx_assert(&sched_lock, MA_OWNED);
705 	/*
706 	 * We need to adjust the nice counts for running KSEs.
707 	 */
708 	if (kg->kg_pri_class == PRI_TIMESHARE)
709 		FOREACH_KSE_IN_GROUP(kg, ke) {
710 			if (ke->ke_state != KES_ONRUNQ &&
711 			    ke->ke_state != KES_THREAD)
712 				continue;
713 			kseq = KSEQ_CPU(ke->ke_cpu);
714 			kseq_nice_rem(kseq, kg->kg_nice);
715 			kseq_nice_add(kseq, nice);
716 		}
717 	kg->kg_nice = nice;
718 	sched_priority(kg);
719 	FOREACH_THREAD_IN_GROUP(kg, td)
720 		td->td_flags |= TDF_NEEDRESCHED;
721 }
722 
723 void
724 sched_sleep(struct thread *td, u_char prio)
725 {
726 	mtx_assert(&sched_lock, MA_OWNED);
727 
728 	td->td_slptime = ticks;
729 	td->td_priority = prio;
730 
731 	CTR2(KTR_ULE, "sleep kse %p (tick: %d)",
732 	    td->td_kse, td->td_slptime);
733 }
734 
735 void
736 sched_wakeup(struct thread *td)
737 {
738 	mtx_assert(&sched_lock, MA_OWNED);
739 
740 	/*
741 	 * Let the kseg know how long we slept for.  This is because process
742 	 * interactivity behavior is modeled in the kseg.
743 	 */
744 	if (td->td_slptime) {
745 		struct ksegrp *kg;
746 		int hzticks;
747 
748 		kg = td->td_ksegrp;
749 		hzticks = ticks - td->td_slptime;
750 		kg->kg_slptime += hzticks << 10;
751 		sched_priority(kg);
752 		CTR2(KTR_ULE, "wakeup kse %p (%d ticks)",
753 		    td->td_kse, hzticks);
754 		td->td_slptime = 0;
755 	}
756 	setrunqueue(td);
757         if (td->td_priority < curthread->td_priority)
758                 curthread->td_flags |= TDF_NEEDRESCHED;
759 }
760 
761 /*
762  * Penalize the parent for creating a new child and initialize the child's
763  * priority.
764  */
765 void
766 sched_fork(struct proc *p, struct proc *p1)
767 {
768 
769 	mtx_assert(&sched_lock, MA_OWNED);
770 
771 	sched_fork_ksegrp(FIRST_KSEGRP_IN_PROC(p), FIRST_KSEGRP_IN_PROC(p1));
772 	sched_fork_kse(FIRST_KSE_IN_PROC(p), FIRST_KSE_IN_PROC(p1));
773 	sched_fork_thread(FIRST_THREAD_IN_PROC(p), FIRST_THREAD_IN_PROC(p1));
774 }
775 
776 void
777 sched_fork_kse(struct kse *ke, struct kse *child)
778 {
779 
780 	child->ke_slice = ke->ke_slice;
781 	child->ke_cpu = ke->ke_cpu; /* sched_pickcpu(); */
782 	child->ke_runq = NULL;
783 
784 	/*
785 	 * Claim that we've been running for one second for statistical
786 	 * purposes.
787 	 */
788 	child->ke_ticks = 0;
789 	child->ke_ltick = ticks;
790 	child->ke_ftick = ticks - hz;
791 }
792 
793 void
794 sched_fork_ksegrp(struct ksegrp *kg, struct ksegrp *child)
795 {
796 
797 	PROC_LOCK_ASSERT(child->kg_proc, MA_OWNED);
798 	/* XXX Need something better here */
799 	if (kg->kg_slptime > kg->kg_runtime) {
800 		child->kg_slptime = SCHED_DYN_RANGE;
801 		child->kg_runtime = kg->kg_slptime / SCHED_DYN_RANGE;
802 	} else {
803 		child->kg_runtime = SCHED_DYN_RANGE;
804 		child->kg_slptime = kg->kg_runtime / SCHED_DYN_RANGE;
805 	}
806 
807 	child->kg_user_pri = kg->kg_user_pri;
808 	child->kg_nice = kg->kg_nice;
809 }
810 
811 void
812 sched_fork_thread(struct thread *td, struct thread *child)
813 {
814 }
815 
816 void
817 sched_class(struct ksegrp *kg, int class)
818 {
819 	struct kseq *kseq;
820 	struct kse *ke;
821 
822 	mtx_assert(&sched_lock, MA_OWNED);
823 	if (kg->kg_pri_class == class)
824 		return;
825 
826 	FOREACH_KSE_IN_GROUP(kg, ke) {
827 		if (ke->ke_state != KES_ONRUNQ &&
828 		    ke->ke_state != KES_THREAD)
829 			continue;
830 		kseq = KSEQ_CPU(ke->ke_cpu);
831 
832 		kseq->ksq_loads[PRI_BASE(kg->kg_pri_class)]--;
833 		kseq->ksq_loads[PRI_BASE(class)]++;
834 
835 		if (kg->kg_pri_class == PRI_TIMESHARE)
836 			kseq_nice_rem(kseq, kg->kg_nice);
837 		else if (class == PRI_TIMESHARE)
838 			kseq_nice_add(kseq, kg->kg_nice);
839 	}
840 
841 	kg->kg_pri_class = class;
842 }
843 
844 /*
845  * Return some of the child's priority and interactivity to the parent.
846  */
847 void
848 sched_exit(struct proc *p, struct proc *child)
849 {
850 	/* XXX Need something better here */
851 	mtx_assert(&sched_lock, MA_OWNED);
852 	sched_exit_kse(FIRST_KSE_IN_PROC(p), FIRST_KSE_IN_PROC(child));
853 }
854 
855 void
856 sched_exit_kse(struct kse *ke, struct kse *child)
857 {
858 	kseq_rem(KSEQ_CPU(child->ke_cpu), child);
859 }
860 
861 void
862 sched_exit_ksegrp(struct ksegrp *kg, struct ksegrp *child)
863 {
864 }
865 
866 void
867 sched_exit_thread(struct thread *td, struct thread *child)
868 {
869 }
870 
871 void
872 sched_clock(struct kse *ke)
873 {
874 	struct kseq *kseq;
875 	struct ksegrp *kg;
876 	struct thread *td;
877 #if 0
878 	struct kse *nke;
879 #endif
880 
881 	/*
882 	 * sched_setup() apparently happens prior to stathz being set.  We
883 	 * need to resolve the timers earlier in the boot so we can avoid
884 	 * calculating this here.
885 	 */
886 	if (realstathz == 0) {
887 		realstathz = stathz ? stathz : hz;
888 		tickincr = hz / realstathz;
889 		/*
890 		 * XXX This does not work for values of stathz that are much
891 		 * larger than hz.
892 		 */
893 		if (tickincr == 0)
894 			tickincr = 1;
895 	}
896 
897 	td = ke->ke_thread;
898 	kg = ke->ke_ksegrp;
899 
900 	mtx_assert(&sched_lock, MA_OWNED);
901 	KASSERT((td != NULL), ("schedclock: null thread pointer"));
902 
903 	/* Adjust ticks for pctcpu */
904 	ke->ke_ticks++;
905 	ke->ke_ltick = ticks;
906 
907 	/* Go up to one second beyond our max and then trim back down */
908 	if (ke->ke_ftick + SCHED_CPU_TICKS + hz < ke->ke_ltick)
909 		sched_pctcpu_update(ke);
910 
911 	if (td->td_flags & TDF_IDLETD)
912 		return;
913 
914 	CTR4(KTR_ULE, "Tick kse %p (slice: %d, slptime: %d, runtime: %d)",
915 	    ke, ke->ke_slice, kg->kg_slptime >> 10, kg->kg_runtime >> 10);
916 
917 	/*
918 	 * We only do slicing code for TIMESHARE ksegrps.
919 	 */
920 	if (kg->kg_pri_class != PRI_TIMESHARE)
921 		return;
922 	/*
923 	 * Check for a higher priority task on the run queue.  This can happen
924 	 * on SMP if another processor woke up a process on our runq.
925 	 */
926 	kseq = KSEQ_SELF();
927 #if 0
928 	if (kseq->ksq_load > 1 && (nke = kseq_choose(kseq)) != NULL) {
929 		if (sched_strict &&
930 		    nke->ke_thread->td_priority < td->td_priority)
931 			td->td_flags |= TDF_NEEDRESCHED;
932 		else if (nke->ke_thread->td_priority <
933 		    td->td_priority SCHED_PRIO_SLOP)
934 
935 		if (nke->ke_thread->td_priority < td->td_priority)
936 			td->td_flags |= TDF_NEEDRESCHED;
937 	}
938 #endif
939 	/*
940 	 * We used a tick charge it to the ksegrp so that we can compute our
941 	 * interactivity.
942 	 */
943 	kg->kg_runtime += tickincr << 10;
944 
945 	/*
946 	 * We used up one time slice.
947 	 */
948 	ke->ke_slice--;
949 #ifdef SMP
950 	kseq->ksq_rslices--;
951 #endif
952 
953 	if (ke->ke_slice > 0)
954 		return;
955 	/*
956 	 * We're out of time, recompute priorities and requeue.
957 	 */
958 	kseq_rem(kseq, ke);
959 	sched_priority(kg);
960 	sched_slice(ke);
961 	if (SCHED_CURR(kg, ke))
962 		ke->ke_runq = kseq->ksq_curr;
963 	else
964 		ke->ke_runq = kseq->ksq_next;
965 	kseq_add(kseq, ke);
966 	td->td_flags |= TDF_NEEDRESCHED;
967 }
968 
969 int
970 sched_runnable(void)
971 {
972 	struct kseq *kseq;
973 	int load;
974 
975 	load = 1;
976 
977 	mtx_lock_spin(&sched_lock);
978 	kseq = KSEQ_SELF();
979 
980 	if (kseq->ksq_load)
981 		goto out;
982 #ifdef SMP
983 	/*
984 	 * For SMP we may steal other processor's KSEs.  Just search until we
985 	 * verify that at least on other cpu has a runnable task.
986 	 */
987 	if (smp_started) {
988 		int i;
989 
990 		for (i = 0; i < mp_maxid; i++) {
991 			if (CPU_ABSENT(i))
992 				continue;
993 			kseq = KSEQ_CPU(i);
994 			if (kseq->ksq_load > 1)
995 				goto out;
996 		}
997 	}
998 #endif
999 	load = 0;
1000 out:
1001 	mtx_unlock_spin(&sched_lock);
1002 	return (load);
1003 }
1004 
1005 void
1006 sched_userret(struct thread *td)
1007 {
1008 	struct ksegrp *kg;
1009 
1010 	kg = td->td_ksegrp;
1011 
1012 	if (td->td_priority != kg->kg_user_pri) {
1013 		mtx_lock_spin(&sched_lock);
1014 		td->td_priority = kg->kg_user_pri;
1015 		mtx_unlock_spin(&sched_lock);
1016 	}
1017 }
1018 
1019 struct kse *
1020 sched_choose(void)
1021 {
1022 	struct kseq *kseq;
1023 	struct kse *ke;
1024 
1025 	mtx_assert(&sched_lock, MA_OWNED);
1026 #ifdef SMP
1027 retry:
1028 #endif
1029 	kseq = KSEQ_SELF();
1030 	ke = kseq_choose(kseq);
1031 	if (ke) {
1032 		runq_remove(ke->ke_runq, ke);
1033 		ke->ke_state = KES_THREAD;
1034 
1035 		if (ke->ke_ksegrp->kg_pri_class == PRI_TIMESHARE) {
1036 			CTR4(KTR_ULE, "Run kse %p from %p (slice: %d, pri: %d)",
1037 			    ke, ke->ke_runq, ke->ke_slice,
1038 			    ke->ke_thread->td_priority);
1039 		}
1040 		return (ke);
1041 	}
1042 
1043 #ifdef SMP
1044 	if (smp_started) {
1045 		/*
1046 		 * Find the cpu with the highest load and steal one proc.
1047 		 */
1048 		if ((kseq = kseq_load_highest()) == NULL)
1049 			return (NULL);
1050 
1051 		/*
1052 		 * Remove this kse from this kseq and runq and then requeue
1053 		 * on the current processor.  Then we will dequeue it
1054 		 * normally above.
1055 		 */
1056 		ke = kseq_choose(kseq);
1057 		runq_remove(ke->ke_runq, ke);
1058 		ke->ke_state = KES_THREAD;
1059 		kseq_rem(kseq, ke);
1060 
1061 		ke->ke_cpu = PCPU_GET(cpuid);
1062 		sched_add(ke);
1063 		goto retry;
1064 	}
1065 #endif
1066 
1067 	return (NULL);
1068 }
1069 
1070 void
1071 sched_add(struct kse *ke)
1072 {
1073 	struct kseq *kseq;
1074 	struct ksegrp *kg;
1075 
1076 	mtx_assert(&sched_lock, MA_OWNED);
1077 	KASSERT((ke->ke_thread != NULL), ("sched_add: No thread on KSE"));
1078 	KASSERT((ke->ke_thread->td_kse != NULL),
1079 	    ("sched_add: No KSE on thread"));
1080 	KASSERT(ke->ke_state != KES_ONRUNQ,
1081 	    ("sched_add: kse %p (%s) already in run queue", ke,
1082 	    ke->ke_proc->p_comm));
1083 	KASSERT(ke->ke_proc->p_sflag & PS_INMEM,
1084 	    ("sched_add: process swapped out"));
1085 	KASSERT(ke->ke_runq == NULL,
1086 	    ("sched_add: KSE %p is still assigned to a run queue", ke));
1087 
1088 	kg = ke->ke_ksegrp;
1089 
1090 	switch (PRI_BASE(kg->kg_pri_class)) {
1091 	case PRI_ITHD:
1092 	case PRI_REALTIME:
1093 		kseq = KSEQ_SELF();
1094 		ke->ke_runq = kseq->ksq_curr;
1095 		ke->ke_slice = SCHED_SLICE_MAX;
1096 		ke->ke_cpu = PCPU_GET(cpuid);
1097 		break;
1098 	case PRI_TIMESHARE:
1099 		kseq = KSEQ_CPU(ke->ke_cpu);
1100 		if (SCHED_CURR(kg, ke))
1101 			ke->ke_runq = kseq->ksq_curr;
1102 		else
1103 			ke->ke_runq = kseq->ksq_next;
1104 		break;
1105 	case PRI_IDLE:
1106 		kseq = KSEQ_CPU(ke->ke_cpu);
1107 		/*
1108 		 * This is for priority prop.
1109 		 */
1110 		if (ke->ke_thread->td_priority < PRI_MAX_TIMESHARE)
1111 			ke->ke_runq = kseq->ksq_curr;
1112 		else
1113 			ke->ke_runq = &kseq->ksq_idle;
1114 		ke->ke_slice = SCHED_SLICE_MIN;
1115 		break;
1116 	default:
1117 		panic("Unknown pri class.\n");
1118 		break;
1119 	}
1120 
1121 	ke->ke_ksegrp->kg_runq_kses++;
1122 	ke->ke_state = KES_ONRUNQ;
1123 
1124 	runq_add(ke->ke_runq, ke);
1125 	kseq_add(kseq, ke);
1126 }
1127 
1128 void
1129 sched_rem(struct kse *ke)
1130 {
1131 	struct kseq *kseq;
1132 
1133 	mtx_assert(&sched_lock, MA_OWNED);
1134 	KASSERT((ke->ke_state == KES_ONRUNQ), ("KSE not on run queue"));
1135 
1136 	ke->ke_state = KES_THREAD;
1137 	ke->ke_ksegrp->kg_runq_kses--;
1138 	kseq = KSEQ_CPU(ke->ke_cpu);
1139 	runq_remove(ke->ke_runq, ke);
1140 	kseq_rem(kseq, ke);
1141 }
1142 
1143 fixpt_t
1144 sched_pctcpu(struct kse *ke)
1145 {
1146 	fixpt_t pctcpu;
1147 
1148 	pctcpu = 0;
1149 
1150 	mtx_lock_spin(&sched_lock);
1151 	if (ke->ke_ticks) {
1152 		int rtick;
1153 
1154 		/* Update to account for time potentially spent sleeping */
1155 		ke->ke_ltick = ticks;
1156 		sched_pctcpu_update(ke);
1157 
1158 		/* How many rtick per second ? */
1159 		rtick = ke->ke_ticks / SCHED_CPU_TIME;
1160 		pctcpu = (FSCALE * ((FSCALE * rtick)/realstathz)) >> FSHIFT;
1161 	}
1162 
1163 	ke->ke_proc->p_swtime = ke->ke_ltick - ke->ke_ftick;
1164 	mtx_unlock_spin(&sched_lock);
1165 
1166 	return (pctcpu);
1167 }
1168 
1169 int
1170 sched_sizeof_kse(void)
1171 {
1172 	return (sizeof(struct kse) + sizeof(struct ke_sched));
1173 }
1174 
1175 int
1176 sched_sizeof_ksegrp(void)
1177 {
1178 	return (sizeof(struct ksegrp) + sizeof(struct kg_sched));
1179 }
1180 
1181 int
1182 sched_sizeof_proc(void)
1183 {
1184 	return (sizeof(struct proc));
1185 }
1186 
1187 int
1188 sched_sizeof_thread(void)
1189 {
1190 	return (sizeof(struct thread) + sizeof(struct td_sched));
1191 }
1192