summaryrefslogtreecommitdiffstats
path: root/sys/kern/kern_synch.c
blob: 24da6e919f168685afdf765300f2b74dddaf74d5 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
/*-
 * Copyright (c) 1982, 1986, 1990, 1991, 1993
 *	The Regents of the University of California.  All rights reserved.
 * (c) UNIX System Laboratories, Inc.
 * All or some portions of this file are derived from material licensed
 * to the University of California by American Telephone and Telegraph
 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
 * the permission of UNIX System Laboratories, Inc.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 3. All advertising materials mentioning features or use of this software
 *    must display the following acknowledgement:
 *	This product includes software developed by the University of
 *	California, Berkeley and its contributors.
 * 4. Neither the name of the University nor the names of its contributors
 *    may be used to endorse or promote products derived from this software
 *    without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 *
 *	@(#)kern_synch.c	8.9 (Berkeley) 5/19/95
 * $FreeBSD$
 */

#include "opt_ddb.h"
#include "opt_ktrace.h"

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/condvar.h>
#include <sys/kernel.h>
#include <sys/ktr.h>
#include <sys/lock.h>
#include <sys/mutex.h>
#include <sys/proc.h>
#include <sys/resourcevar.h>
#include <sys/signalvar.h>
#include <sys/smp.h>
#include <sys/sx.h>
#include <sys/sysctl.h>
#include <sys/sysproto.h>
#include <sys/vmmeter.h>
#include <vm/vm.h>
#include <vm/vm_extern.h>
#ifdef DDB
#include <ddb/ddb.h>
#endif
#ifdef KTRACE
#include <sys/uio.h>
#include <sys/ktrace.h>
#endif

#include <machine/cpu.h>

static void sched_setup __P((void *dummy));
SYSINIT(sched_setup, SI_SUB_KICK_SCHEDULER, SI_ORDER_FIRST, sched_setup, NULL)

int	hogticks;
int	lbolt;
int	sched_quantum;		/* Roundrobin scheduling quantum in ticks. */

static struct callout schedcpu_callout;
static struct callout roundrobin_callout;

static void	endtsleep __P((void *));
static void	roundrobin __P((void *arg));
static void	schedcpu __P((void *arg));

static int
sysctl_kern_quantum(SYSCTL_HANDLER_ARGS)
{
	int error, new_val;

	new_val = sched_quantum * tick;
	error = sysctl_handle_int(oidp, &new_val, 0, req);
        if (error != 0 || req->newptr == NULL)
		return (error);
	if (new_val < tick)
		return (EINVAL);
	sched_quantum = new_val / tick;
	hogticks = 2 * sched_quantum;
	return (0);
}

SYSCTL_PROC(_kern, OID_AUTO, quantum, CTLTYPE_INT|CTLFLAG_RW,
	0, sizeof sched_quantum, sysctl_kern_quantum, "I", "");

/*
 * Arrange to reschedule if necessary, taking the priorities and
 * schedulers into account.
 */
void
maybe_resched(p)
	struct proc *p;
{

	mtx_assert(&sched_lock, MA_OWNED);
	if (p->p_pri.pri_level < curproc->p_pri.pri_level)
		curproc->p_sflag |= PS_NEEDRESCHED;
}

int 
roundrobin_interval(void)
{
	return (sched_quantum);
}

/*
 * Force switch among equal priority processes every 100ms.
 * We don't actually need to force a context switch of the current process.
 * The act of firing the event triggers a context switch to softclock() and
 * then switching back out again which is equivalent to a preemption, thus
 * no further work is needed on the local CPU.
 */
/* ARGSUSED */
static void
roundrobin(arg)
	void *arg;
{

#ifdef SMP
	mtx_lock_spin(&sched_lock);
	forward_roundrobin();
	mtx_unlock_spin(&sched_lock);
#endif

	callout_reset(&roundrobin_callout, sched_quantum, roundrobin, NULL);
}

/*
 * Constants for digital decay and forget:
 *	90% of (p_estcpu) usage in 5 * loadav time
 *	95% of (p_pctcpu) usage in 60 seconds (load insensitive)
 *          Note that, as ps(1) mentions, this can let percentages
 *          total over 100% (I've seen 137.9% for 3 processes).
 *
 * Note that schedclock() updates p_estcpu and p_cpticks asynchronously.
 *
 * We wish to decay away 90% of p_estcpu in (5 * loadavg) seconds.
 * That is, the system wants to compute a value of decay such
 * that the following for loop:
 * 	for (i = 0; i < (5 * loadavg); i++)
 * 		p_estcpu *= decay;
 * will compute
 * 	p_estcpu *= 0.1;
 * for all values of loadavg:
 *
 * Mathematically this loop can be expressed by saying:
 * 	decay ** (5 * loadavg) ~= .1
 *
 * The system computes decay as:
 * 	decay = (2 * loadavg) / (2 * loadavg + 1)
 *
 * We wish to prove that the system's computation of decay
 * will always fulfill the equation:
 * 	decay ** (5 * loadavg) ~= .1
 *
 * If we compute b as:
 * 	b = 2 * loadavg
 * then
 * 	decay = b / (b + 1)
 *
 * We now need to prove two things:
 *	1) Given factor ** (5 * loadavg) ~= .1, prove factor == b/(b+1)
 *	2) Given b/(b+1) ** power ~= .1, prove power == (5 * loadavg)
 *
 * Facts:
 *         For x close to zero, exp(x) =~ 1 + x, since
 *              exp(x) = 0! + x**1/1! + x**2/2! + ... .
 *              therefore exp(-1/b) =~ 1 - (1/b) = (b-1)/b.
 *         For x close to zero, ln(1+x) =~ x, since
 *              ln(1+x) = x - x**2/2 + x**3/3 - ...     -1 < x < 1
 *              therefore ln(b/(b+1)) = ln(1 - 1/(b+1)) =~ -1/(b+1).
 *         ln(.1) =~ -2.30
 *
 * Proof of (1):
 *    Solve (factor)**(power) =~ .1 given power (5*loadav):
 *	solving for factor,
 *      ln(factor) =~ (-2.30/5*loadav), or
 *      factor =~ exp(-1/((5/2.30)*loadav)) =~ exp(-1/(2*loadav)) =
 *          exp(-1/b) =~ (b-1)/b =~ b/(b+1).                    QED
 *
 * Proof of (2):
 *    Solve (factor)**(power) =~ .1 given factor == (b/(b+1)):
 *	solving for power,
 *      power*ln(b/(b+1)) =~ -2.30, or
 *      power =~ 2.3 * (b + 1) = 4.6*loadav + 2.3 =~ 5*loadav.  QED
 *
 * Actual power values for the implemented algorithm are as follows:
 *      loadav: 1       2       3       4
 *      power:  5.68    10.32   14.94   19.55
 */

/* calculations for digital decay to forget 90% of usage in 5*loadav sec */
#define	loadfactor(loadav)	(2 * (loadav))
#define	decay_cpu(loadfac, cpu)	(((loadfac) * (cpu)) / ((loadfac) + FSCALE))

/* decay 95% of `p_pctcpu' in 60 seconds; see CCPU_SHIFT before changing */
static fixpt_t	ccpu = 0.95122942450071400909 * FSCALE;	/* exp(-1/20) */
SYSCTL_INT(_kern, OID_AUTO, ccpu, CTLFLAG_RD, &ccpu, 0, "");

/* kernel uses `FSCALE', userland (SHOULD) use kern.fscale */
static int	fscale __unused = FSCALE;
SYSCTL_INT(_kern, OID_AUTO, fscale, CTLFLAG_RD, 0, FSCALE, "");

/*
 * If `ccpu' is not equal to `exp(-1/20)' and you still want to use the
 * faster/more-accurate formula, you'll have to estimate CCPU_SHIFT below
 * and possibly adjust FSHIFT in "param.h" so that (FSHIFT >= CCPU_SHIFT).
 *
 * To estimate CCPU_SHIFT for exp(-1/20), the following formula was used:
 *	1 - exp(-1/20) ~= 0.0487 ~= 0.0488 == 1 (fixed pt, *11* bits).
 *
 * If you don't want to bother with the faster/more-accurate formula, you
 * can set CCPU_SHIFT to (FSHIFT + 1) which will use a slower/less-accurate
 * (more general) method of calculating the %age of CPU used by a process.
 */
#define	CCPU_SHIFT	11

/*
 * Recompute process priorities, every hz ticks.
 * MP-safe, called without the Giant mutex.
 */
/* ARGSUSED */
static void
schedcpu(arg)
	void *arg;
{
	register fixpt_t loadfac = loadfactor(averunnable.ldavg[0]);
	register struct proc *p;
	register int realstathz;

	realstathz = stathz ? stathz : hz;
	sx_slock(&allproc_lock);
	LIST_FOREACH(p, &allproc, p_list) {
		/*
		 * Increment time in/out of memory and sleep time
		 * (if sleeping).  We ignore overflow; with 16-bit int's
		 * (remember them?) overflow takes 45 days.
		 */
		mtx_lock_spin(&sched_lock);
		p->p_swtime++;
		if (p->p_stat == SSLEEP || p->p_stat == SSTOP)
			p->p_slptime++;
		p->p_pctcpu = (p->p_pctcpu * ccpu) >> FSHIFT;
		/*
		 * If the process has slept the entire second,
		 * stop recalculating its priority until it wakes up.
		 */
		if (p->p_slptime > 1) {
			mtx_unlock_spin(&sched_lock);
			continue;
		}

		/*
		 * p_pctcpu is only for ps.
		 */
#if	(FSHIFT >= CCPU_SHIFT)
		p->p_pctcpu += (realstathz == 100)?
			((fixpt_t) p->p_cpticks) << (FSHIFT - CCPU_SHIFT):
                	100 * (((fixpt_t) p->p_cpticks)
				<< (FSHIFT - CCPU_SHIFT)) / realstathz;
#else
		p->p_pctcpu += ((FSCALE - ccpu) *
			(p->p_cpticks * FSCALE / realstathz)) >> FSHIFT;
#endif
		p->p_cpticks = 0;
		p->p_estcpu = decay_cpu(loadfac, p->p_estcpu);
		resetpriority(p);
		if (p->p_pri.pri_level >= PUSER) {
			if (p->p_oncpu == NOCPU && 	/* idle */
			    p->p_stat == SRUN &&
			    (p->p_sflag & PS_INMEM) &&
			    (p->p_pri.pri_level / RQ_PPQ) !=
			    (p->p_pri.pri_user / RQ_PPQ)) {
				remrunqueue(p);
				p->p_pri.pri_level = p->p_pri.pri_user;
				setrunqueue(p);
			} else
				p->p_pri.pri_level = p->p_pri.pri_user;
		}
		mtx_unlock_spin(&sched_lock);
	}
	sx_sunlock(&allproc_lock);
	vmmeter();
	wakeup((caddr_t)&lbolt);
	callout_reset(&schedcpu_callout, hz, schedcpu, NULL);
}

/*
 * Recalculate the priority of a process after it has slept for a while.
 * For all load averages >= 1 and max p_estcpu of 255, sleeping for at
 * least six times the loadfactor will decay p_estcpu to zero.
 */
void
updatepri(p)
	register struct proc *p;
{
	register unsigned int newcpu = p->p_estcpu;
	register fixpt_t loadfac = loadfactor(averunnable.ldavg[0]);

	if (p->p_slptime > 5 * loadfac)
		p->p_estcpu = 0;
	else {
		p->p_slptime--;	/* the first time was done in schedcpu */
		while (newcpu && --p->p_slptime)
			newcpu = decay_cpu(loadfac, newcpu);
		p->p_estcpu = newcpu;
	}
	resetpriority(p);
}

/*
 * We're only looking at 7 bits of the address; everything is
 * aligned to 4, lots of things are aligned to greater powers
 * of 2.  Shift right by 8, i.e. drop the bottom 256 worth.
 */
#define TABLESIZE	128
static TAILQ_HEAD(slpquehead, proc) slpque[TABLESIZE];
#define LOOKUP(x)	(((intptr_t)(x) >> 8) & (TABLESIZE - 1))

void
sleepinit(void)
{
	int i;

	sched_quantum = hz/10;
	hogticks = 2 * sched_quantum;
	for (i = 0; i < TABLESIZE; i++)
		TAILQ_INIT(&slpque[i]);
}

/*
 * General sleep call.  Suspends the current process until a wakeup is
 * performed on the specified identifier.  The process will then be made
 * runnable with the specified priority.  Sleeps at most timo/hz seconds
 * (0 means no timeout).  If pri includes PCATCH flag, signals are checked
 * before and after sleeping, else signals are not checked.  Returns 0 if
 * awakened, EWOULDBLOCK if the timeout expires.  If PCATCH is set and a
 * signal needs to be delivered, ERESTART is returned if the current system
 * call should be restarted if possible, and EINTR is returned if the system
 * call should be interrupted by the signal (return EINTR).
 *
 * The mutex argument is exited before the caller is suspended, and
 * entered before msleep returns.  If priority includes the PDROP
 * flag the mutex is not entered before returning.
 */
int
msleep(ident, mtx, priority, wmesg, timo)
	void *ident;
	struct mtx *mtx;
	int priority, timo;
	const char *wmesg;
{
	struct proc *p = curproc;
	int sig, catch = priority & PCATCH;
	int rval = 0;
	WITNESS_SAVE_DECL(mtx);

#ifdef KTRACE
	if (p && KTRPOINT(p, KTR_CSW))
		ktrcsw(p->p_tracep, 1, 0);
#endif
	WITNESS_SLEEP(0, &mtx->mtx_object);
	KASSERT(timo != 0 || mtx_owned(&Giant) || mtx != NULL,
	    ("sleeping without a mutex"));
	mtx_lock_spin(&sched_lock);
	if (cold || panicstr) {
		/*
		 * After a panic, or during autoconfiguration,
		 * just give interrupts a chance, then just return;
		 * don't run any other procs or panic below,
		 * in case this is the idle process and already asleep.
		 */
		if (mtx != NULL && priority & PDROP)
			mtx_unlock_flags(mtx, MTX_NOSWITCH);
		mtx_unlock_spin(&sched_lock);
		return (0);
	}

	DROP_GIANT_NOSWITCH();

	if (mtx != NULL) {
		mtx_assert(mtx, MA_OWNED | MA_NOTRECURSED);
		WITNESS_SAVE(&mtx->mtx_object, mtx);
		mtx_unlock_flags(mtx, MTX_NOSWITCH);
		if (priority & PDROP)
			mtx = NULL;
	}

	KASSERT(p != NULL, ("msleep1"));
	KASSERT(ident != NULL && p->p_stat == SRUN, ("msleep"));

	p->p_wchan = ident;
	p->p_wmesg = wmesg;
	p->p_slptime = 0;
	p->p_pri.pri_level = priority & PRIMASK;
	CTR5(KTR_PROC, "msleep: proc %p (pid %d, %s) on %s (%p)", p, p->p_pid,
	    p->p_comm, wmesg, ident);
	TAILQ_INSERT_TAIL(&slpque[LOOKUP(ident)], p, p_slpq);
	if (timo)
		callout_reset(&p->p_slpcallout, timo, endtsleep, p);
	/*
	 * We put ourselves on the sleep queue and start our timeout
	 * before calling CURSIG, as we could stop there, and a wakeup
	 * or a SIGCONT (or both) could occur while we were stopped.
	 * A SIGCONT would cause us to be marked as SSLEEP
	 * without resuming us, thus we must be ready for sleep
	 * when CURSIG is called.  If the wakeup happens while we're
	 * stopped, p->p_wchan will be 0 upon return from CURSIG.
	 */
	if (catch) {
		CTR3(KTR_PROC, "msleep caught: proc %p (pid %d, %s)", p,
		    p->p_pid, p->p_comm);
		p->p_sflag |= PS_SINTR;
		mtx_unlock_spin(&sched_lock);
		PROC_LOCK(p);
		sig = CURSIG(p);
		mtx_lock_spin(&sched_lock);
		PROC_UNLOCK_NOSWITCH(p);
		if (sig != 0) {
			if (p->p_wchan != NULL)
				unsleep(p);
		} else if (p->p_wchan == NULL)
			catch = 0;
	} else
		sig = 0;
	if (p->p_wchan != NULL) {
		p->p_stat = SSLEEP;
		p->p_stats->p_ru.ru_nvcsw++;
		mi_switch();
	}
	CTR3(KTR_PROC, "msleep resume: proc %p (pid %d, %s)", p, p->p_pid,
	    p->p_comm);
	KASSERT(p->p_stat == SRUN, ("running but not SRUN"));
	p->p_sflag &= ~PS_SINTR;
	if (p->p_sflag & PS_TIMEOUT) {
		p->p_sflag &= ~PS_TIMEOUT;
		if (sig == 0)
			rval = EWOULDBLOCK;
	} else if (p->p_sflag & PS_TIMOFAIL)
		p->p_sflag &= ~PS_TIMOFAIL;
	else if (timo && callout_stop(&p->p_slpcallout) == 0) {
		/*
		 * This isn't supposed to be pretty.  If we are here, then
		 * the endtsleep() callout is currently executing on another
		 * CPU and is either spinning on the sched_lock or will be
		 * soon.  If we don't synchronize here, there is a chance
		 * that this process may msleep() again before the callout
		 * has a chance to run and the callout may end up waking up
		 * the wrong msleep().  Yuck.
		 */
		p->p_sflag |= PS_TIMEOUT;
		p->p_stats->p_ru.ru_nivcsw++;
		mi_switch();
	}
	mtx_unlock_spin(&sched_lock);

	if (rval == 0 && catch) {
		PROC_LOCK(p);
		/* XXX: shouldn't we always be calling CURSIG() */ 
		if (sig != 0 || (sig = CURSIG(p))) {
			if (SIGISMEMBER(p->p_sigacts->ps_sigintr, sig))
				rval = EINTR;
			else
				rval = ERESTART;
		}
		PROC_UNLOCK(p);
	}
	PICKUP_GIANT();
#ifdef KTRACE
	mtx_lock(&Giant);
	if (KTRPOINT(p, KTR_CSW))
		ktrcsw(p->p_tracep, 0, 0);
	mtx_unlock(&Giant);
#endif
	if (mtx != NULL) {
		mtx_lock(mtx);
		WITNESS_RESTORE(&mtx->mtx_object, mtx);
	}
	return (rval);
}

/*
 * Implement timeout for msleep()
 *
 * If process hasn't been awakened (wchan non-zero),
 * set timeout flag and undo the sleep.  If proc
 * is stopped, just unsleep so it will remain stopped.
 * MP-safe, called without the Giant mutex.
 */
static void
endtsleep(arg)
	void *arg;
{
	register struct proc *p;

	p = (struct proc *)arg;
	CTR3(KTR_PROC, "endtsleep: proc %p (pid %d, %s)", p, p->p_pid,
	    p->p_comm);
	mtx_lock_spin(&sched_lock);
	/*
	 * This is the other half of the synchronization with msleep()
	 * described above.  If the PS_TIMEOUT flag is set, we lost the
	 * race and just need to put the process back on the runqueue.
	 */
	if ((p->p_sflag & PS_TIMEOUT) != 0) {
		p->p_sflag &= ~PS_TIMEOUT;
		setrunqueue(p);
	} else if (p->p_wchan != NULL) {
		if (p->p_stat == SSLEEP)
			setrunnable(p);
		else
			unsleep(p);
		p->p_sflag |= PS_TIMEOUT;
	} else
		p->p_sflag |= PS_TIMOFAIL;
	mtx_unlock_spin(&sched_lock);
}

/*
 * Remove a process from its wait queue
 */
void
unsleep(p)
	register struct proc *p;
{

	mtx_lock_spin(&sched_lock);
	if (p->p_wchan != NULL) {
		TAILQ_REMOVE(&slpque[LOOKUP(p->p_wchan)], p, p_slpq);
		p->p_wchan = NULL;
	}
	mtx_unlock_spin(&sched_lock);
}

/*
 * Make all processes sleeping on the specified identifier runnable.
 */
void
wakeup(ident)
	register void *ident;
{
	register struct slpquehead *qp;
	register struct proc *p;

	mtx_lock_spin(&sched_lock);
	qp = &slpque[LOOKUP(ident)];
restart:
	TAILQ_FOREACH(p, qp, p_slpq) {
		if (p->p_wchan == ident) {
			TAILQ_REMOVE(qp, p, p_slpq);
			p->p_wchan = NULL;
			if (p->p_stat == SSLEEP) {
				/* OPTIMIZED EXPANSION OF setrunnable(p); */
				CTR3(KTR_PROC, "wakeup: proc %p (pid %d, %s)",
				    p, p->p_pid, p->p_comm);
				if (p->p_slptime > 1)
					updatepri(p);
				p->p_slptime = 0;
				p->p_stat = SRUN;
				if (p->p_sflag & PS_INMEM) {
					setrunqueue(p);
					maybe_resched(p);
				} else {
					p->p_sflag |= PS_SWAPINREQ;
					wakeup((caddr_t)&proc0);
				}
				/* END INLINE EXPANSION */
				goto restart;
			}
		}
	}
	mtx_unlock_spin(&sched_lock);
}

/*
 * Make a process sleeping on the specified identifier runnable.
 * May wake more than one process if a target process is currently
 * swapped out.
 */
void
wakeup_one(ident)
	register void *ident;
{
	register struct slpquehead *qp;
	register struct proc *p;

	mtx_lock_spin(&sched_lock);
	qp = &slpque[LOOKUP(ident)];

	TAILQ_FOREACH(p, qp, p_slpq) {
		if (p->p_wchan == ident) {
			TAILQ_REMOVE(qp, p, p_slpq);
			p->p_wchan = NULL;
			if (p->p_stat == SSLEEP) {
				/* OPTIMIZED EXPANSION OF setrunnable(p); */
				CTR3(KTR_PROC, "wakeup1: proc %p (pid %d, %s)",
				    p, p->p_pid, p->p_comm);
				if (p->p_slptime > 1)
					updatepri(p);
				p->p_slptime = 0;
				p->p_stat = SRUN;
				if (p->p_sflag & PS_INMEM) {
					setrunqueue(p);
					maybe_resched(p);
					break;
				} else {
					p->p_sflag |= PS_SWAPINREQ;
					wakeup((caddr_t)&proc0);
				}
				/* END INLINE EXPANSION */
			}
		}
	}
	mtx_unlock_spin(&sched_lock);
}

/*
 * The machine independent parts of mi_switch().
 */
void
mi_switch()
{
	struct timeval new_switchtime;
	register struct proc *p = curproc;	/* XXX */
#if 0
	register struct rlimit *rlim;
#endif
	critical_t sched_crit;
	u_int sched_nest;

	mtx_assert(&sched_lock, MA_OWNED | MA_NOTRECURSED);

	/*
	 * Compute the amount of time during which the current
	 * process was running, and add that to its total so far.
	 */
	microuptime(&new_switchtime);
	if (timevalcmp(&new_switchtime, PCPU_PTR(switchtime), <)) {
#if 0
		/* XXX: This doesn't play well with sched_lock right now. */
		printf("microuptime() went backwards (%ld.%06ld -> %ld.%06ld)\n",
		    PCPU_GET(switchtime.tv_sec), PCPU_GET(switchtime.tv_usec),
		    new_switchtime.tv_sec, new_switchtime.tv_usec);
#endif
		new_switchtime = PCPU_GET(switchtime);
	} else {
		p->p_runtime += (new_switchtime.tv_usec - PCPU_GET(switchtime.tv_usec)) +
		    (new_switchtime.tv_sec - PCPU_GET(switchtime.tv_sec)) *
		    (int64_t)1000000;
	}

#ifdef DDB
	/*
	 * Don't perform context switches from the debugger.
	 */
	if (db_active) {
		mtx_unlock_spin(&sched_lock);
		db_error("Context switches not allowed in the debugger.");
	}
#endif

#if 0
	/*
	 * Check if the process exceeds its cpu resource allocation.
	 * If over max, kill it.
	 *
	 * XXX drop sched_lock, pickup Giant
	 */
	if (p->p_stat != SZOMB && p->p_limit->p_cpulimit != RLIM_INFINITY &&
	    p->p_runtime > p->p_limit->p_cpulimit) {
		rlim = &p->p_rlimit[RLIMIT_CPU];
		if (p->p_runtime / (rlim_t)1000000 >= rlim->rlim_max) {
			mtx_unlock_spin(&sched_lock);
			PROC_LOCK(p);
			killproc(p, "exceeded maximum CPU limit");
			mtx_lock_spin(&sched_lock);
			PROC_UNLOCK_NOSWITCH(p);
		} else {
			mtx_unlock_spin(&sched_lock);
			PROC_LOCK(p);
			psignal(p, SIGXCPU);
			mtx_lock_spin(&sched_lock);
			PROC_UNLOCK_NOSWITCH(p);
			if (rlim->rlim_cur < rlim->rlim_max) {
				/* XXX: we should make a private copy */
				rlim->rlim_cur += 5;
			}
		}
	}
#endif

	/*
	 * Pick a new current process and record its start time.
	 */
	cnt.v_swtch++;
	PCPU_SET(switchtime, new_switchtime);
	CTR3(KTR_PROC, "mi_switch: old proc %p (pid %d, %s)", p, p->p_pid,
	    p->p_comm);
	sched_crit = sched_lock.mtx_savecrit;
	sched_nest = sched_lock.mtx_recurse;
	p->p_lastcpu = p->p_oncpu;
	p->p_oncpu = NOCPU;
	p->p_sflag &= ~PS_NEEDRESCHED;
	cpu_switch();
	p->p_oncpu = PCPU_GET(cpuid);
	sched_lock.mtx_savecrit = sched_crit;
	sched_lock.mtx_recurse = sched_nest;
	sched_lock.mtx_lock = (uintptr_t)p;
	CTR3(KTR_PROC, "mi_switch: new proc %p (pid %d, %s)", p, p->p_pid,
	    p->p_comm);
	if (PCPU_GET(switchtime.tv_sec) == 0)
		microuptime(PCPU_PTR(switchtime));
	PCPU_SET(switchticks, ticks);
}

/*
 * Change process state to be runnable,
 * placing it on the run queue if it is in memory,
 * and awakening the swapper if it isn't in memory.
 */
void
setrunnable(p)
	register struct proc *p;
{

	mtx_lock_spin(&sched_lock);
	switch (p->p_stat) {
	case 0:
	case SRUN:
	case SZOMB:
	case SWAIT:
	default:
		panic("setrunnable");
	case SSTOP:
	case SSLEEP:			/* e.g. when sending signals */
		if (p->p_sflag & PS_CVWAITQ)
			cv_waitq_remove(p);
		else
			unsleep(p);
		break;

	case SIDL:
		break;
	}
	p->p_stat = SRUN;
	if (p->p_slptime > 1)
		updatepri(p);
	p->p_slptime = 0;
	if ((p->p_sflag & PS_INMEM) == 0) {
		p->p_sflag |= PS_SWAPINREQ;
		wakeup((caddr_t)&proc0);
	} else {
		setrunqueue(p);
		maybe_resched(p);
	}
	mtx_unlock_spin(&sched_lock);
}

/*
 * Compute the priority of a process when running in user mode.
 * Arrange to reschedule if the resulting priority is better
 * than that of the current process.
 */
void
resetpriority(p)
	register struct proc *p;
{
	register unsigned int newpriority;

	mtx_lock_spin(&sched_lock);
	if (p->p_pri.pri_class == PRI_TIMESHARE) {
		newpriority = PUSER + p->p_estcpu / INVERSE_ESTCPU_WEIGHT +
		    NICE_WEIGHT * (p->p_nice - PRIO_MIN);
		newpriority = min(max(newpriority, PRI_MIN_TIMESHARE),
		    PRI_MAX_TIMESHARE);
		p->p_pri.pri_user = newpriority;
	}
	maybe_resched(p);
	mtx_unlock_spin(&sched_lock);
}

/* ARGSUSED */
static void
sched_setup(dummy)
	void *dummy;
{

	callout_init(&schedcpu_callout, 1);
	callout_init(&roundrobin_callout, 0);

	/* Kick off timeout driven events by calling first time. */
	roundrobin(NULL);
	schedcpu(NULL);
}

/*
 * We adjust the priority of the current process.  The priority of
 * a process gets worse as it accumulates CPU time.  The cpu usage
 * estimator (p_estcpu) is increased here.  resetpriority() will
 * compute a different priority each time p_estcpu increases by
 * INVERSE_ESTCPU_WEIGHT
 * (until MAXPRI is reached).  The cpu usage estimator ramps up
 * quite quickly when the process is running (linearly), and decays
 * away exponentially, at a rate which is proportionally slower when
 * the system is busy.  The basic principle is that the system will
 * 90% forget that the process used a lot of CPU time in 5 * loadav
 * seconds.  This causes the system to favor processes which haven't
 * run much recently, and to round-robin among other processes.
 */
void
schedclock(p)
	struct proc *p;
{

	p->p_cpticks++;
	p->p_estcpu = ESTCPULIM(p->p_estcpu + 1);
	if ((p->p_estcpu % INVERSE_ESTCPU_WEIGHT) == 0) {
		resetpriority(p);
		if (p->p_pri.pri_level >= PUSER)
			p->p_pri.pri_level = p->p_pri.pri_user;
	}
}

/*
 * General purpose yield system call
 */
int
yield(struct proc *p, struct yield_args *uap)
{
	p->p_retval[0] = 0;

	mtx_lock_spin(&sched_lock);
	mtx_assert(&Giant, MA_NOTOWNED);
#if 0
	DROP_GIANT_NOSWITCH();
#endif
	p->p_pri.pri_level = PRI_MAX_TIMESHARE;
	setrunqueue(p);
	p->p_stats->p_ru.ru_nvcsw++;
	mi_switch();
	mtx_unlock_spin(&sched_lock);
#if 0
	PICKUP_GIANT();
#endif

	return (0);
}

OpenPOWER on IntegriCloud