xref: /netbsd-src/sys/kern/kern_synch.c (revision 5aefcfdc06931dd97e76246d2fe0302f7b3fe094)
1 /*	$NetBSD: kern_synch.c,v 1.99 2000/12/22 22:59:00 jdolecek Exp $	*/
2 
3 /*-
4  * Copyright (c) 1999, 2000 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Jason R. Thorpe of the Numerical Aerospace Simulation Facility,
9  * NASA Ames Research Center.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 3. All advertising materials mentioning features or use of this software
20  *    must display the following acknowledgement:
21  *	This product includes software developed by the NetBSD
22  *	Foundation, Inc. and its contributors.
23  * 4. Neither the name of The NetBSD Foundation nor the names of its
24  *    contributors may be used to endorse or promote products derived
25  *    from this software without specific prior written permission.
26  *
27  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
28  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
29  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
30  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
31  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
32  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
33  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
34  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
35  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
36  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
37  * POSSIBILITY OF SUCH DAMAGE.
38  */
39 
40 /*-
41  * Copyright (c) 1982, 1986, 1990, 1991, 1993
42  *	The Regents of the University of California.  All rights reserved.
43  * (c) UNIX System Laboratories, Inc.
44  * All or some portions of this file are derived from material licensed
45  * to the University of California by American Telephone and Telegraph
46  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
47  * the permission of UNIX System Laboratories, Inc.
48  *
49  * Redistribution and use in source and binary forms, with or without
50  * modification, are permitted provided that the following conditions
51  * are met:
52  * 1. Redistributions of source code must retain the above copyright
53  *    notice, this list of conditions and the following disclaimer.
54  * 2. Redistributions in binary form must reproduce the above copyright
55  *    notice, this list of conditions and the following disclaimer in the
56  *    documentation and/or other materials provided with the distribution.
57  * 3. All advertising materials mentioning features or use of this software
58  *    must display the following acknowledgement:
59  *	This product includes software developed by the University of
60  *	California, Berkeley and its contributors.
61  * 4. Neither the name of the University nor the names of its contributors
62  *    may be used to endorse or promote products derived from this software
63  *    without specific prior written permission.
64  *
65  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
66  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
67  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
68  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
69  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
70  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
71  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
72  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
73  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
74  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
75  * SUCH DAMAGE.
76  *
77  *	@(#)kern_synch.c	8.9 (Berkeley) 5/19/95
78  */
79 
80 #include "opt_ddb.h"
81 #include "opt_ktrace.h"
82 #include "opt_lockdebug.h"
83 #include "opt_multiprocessor.h"
84 
85 #include <sys/param.h>
86 #include <sys/systm.h>
87 #include <sys/callout.h>
88 #include <sys/proc.h>
89 #include <sys/kernel.h>
90 #include <sys/buf.h>
91 #include <sys/signalvar.h>
92 #include <sys/resourcevar.h>
93 #include <sys/sched.h>
94 
95 #include <uvm/uvm_extern.h>
96 
97 #ifdef KTRACE
98 #include <sys/ktrace.h>
99 #endif
100 
101 #include <machine/cpu.h>
102 
103 int	lbolt;			/* once a second sleep address */
104 int	rrticks;		/* number of hardclock ticks per roundrobin() */
105 
106 /*
107  * The global scheduler state.
108  */
109 struct prochd sched_qs[RUNQUE_NQS];	/* run queues */
110 __volatile u_int32_t sched_whichqs;	/* bitmap of non-empty queues */
111 struct slpque sched_slpque[SLPQUE_TABLESIZE]; /* sleep queues */
112 
113 struct simplelock sched_lock = SIMPLELOCK_INITIALIZER;
114 #if defined(MULTIPROCESSOR)
115 struct lock kernel_lock;
116 #endif
117 
118 void schedcpu(void *);
119 void updatepri(struct proc *);
120 void endtsleep(void *);
121 
122 __inline void awaken(struct proc *);
123 
124 struct callout schedcpu_ch = CALLOUT_INITIALIZER;
125 
126 /*
127  * Force switch among equal priority processes every 100ms.
128  * Called from hardclock every hz/10 == rrticks hardclock ticks.
129  */
130 /* ARGSUSED */
131 void
132 roundrobin(struct cpu_info *ci)
133 {
134 	struct schedstate_percpu *spc = &ci->ci_schedstate;
135 
136 	spc->spc_rrticks = rrticks;
137 
138 	if (curproc != NULL) {
139 		if (spc->spc_flags & SPCF_SEENRR) {
140 			/*
141 			 * The process has already been through a roundrobin
142 			 * without switching and may be hogging the CPU.
143 			 * Indicate that the process should yield.
144 			 */
145 			spc->spc_flags |= SPCF_SHOULDYIELD;
146 		} else
147 			spc->spc_flags |= SPCF_SEENRR;
148 	}
149 	need_resched(curcpu());
150 }
151 
152 /*
153  * Constants for digital decay and forget:
154  *	90% of (p_estcpu) usage in 5 * loadav time
155  *	95% of (p_pctcpu) usage in 60 seconds (load insensitive)
156  *          Note that, as ps(1) mentions, this can let percentages
157  *          total over 100% (I've seen 137.9% for 3 processes).
158  *
159  * Note that hardclock updates p_estcpu and p_cpticks independently.
160  *
161  * We wish to decay away 90% of p_estcpu in (5 * loadavg) seconds.
162  * That is, the system wants to compute a value of decay such
163  * that the following for loop:
164  * 	for (i = 0; i < (5 * loadavg); i++)
165  * 		p_estcpu *= decay;
166  * will compute
167  * 	p_estcpu *= 0.1;
168  * for all values of loadavg:
169  *
170  * Mathematically this loop can be expressed by saying:
171  * 	decay ** (5 * loadavg) ~= .1
172  *
173  * The system computes decay as:
174  * 	decay = (2 * loadavg) / (2 * loadavg + 1)
175  *
176  * We wish to prove that the system's computation of decay
177  * will always fulfill the equation:
178  * 	decay ** (5 * loadavg) ~= .1
179  *
180  * If we compute b as:
181  * 	b = 2 * loadavg
182  * then
183  * 	decay = b / (b + 1)
184  *
185  * We now need to prove two things:
186  *	1) Given factor ** (5 * loadavg) ~= .1, prove factor == b/(b+1)
187  *	2) Given b/(b+1) ** power ~= .1, prove power == (5 * loadavg)
188  *
189  * Facts:
190  *         For x close to zero, exp(x) =~ 1 + x, since
191  *              exp(x) = 0! + x**1/1! + x**2/2! + ... .
192  *              therefore exp(-1/b) =~ 1 - (1/b) = (b-1)/b.
193  *         For x close to zero, ln(1+x) =~ x, since
194  *              ln(1+x) = x - x**2/2 + x**3/3 - ...     -1 < x < 1
195  *              therefore ln(b/(b+1)) = ln(1 - 1/(b+1)) =~ -1/(b+1).
196  *         ln(.1) =~ -2.30
197  *
198  * Proof of (1):
199  *    Solve (factor)**(power) =~ .1 given power (5*loadav):
200  *	solving for factor,
201  *      ln(factor) =~ (-2.30/5*loadav), or
202  *      factor =~ exp(-1/((5/2.30)*loadav)) =~ exp(-1/(2*loadav)) =
203  *          exp(-1/b) =~ (b-1)/b =~ b/(b+1).                    QED
204  *
205  * Proof of (2):
206  *    Solve (factor)**(power) =~ .1 given factor == (b/(b+1)):
207  *	solving for power,
208  *      power*ln(b/(b+1)) =~ -2.30, or
209  *      power =~ 2.3 * (b + 1) = 4.6*loadav + 2.3 =~ 5*loadav.  QED
210  *
211  * Actual power values for the implemented algorithm are as follows:
212  *      loadav: 1       2       3       4
213  *      power:  5.68    10.32   14.94   19.55
214  */
215 
216 /* calculations for digital decay to forget 90% of usage in 5*loadav sec */
217 #define	loadfactor(loadav)	(2 * (loadav))
218 #define	decay_cpu(loadfac, cpu)	(((loadfac) * (cpu)) / ((loadfac) + FSCALE))
219 
220 /* decay 95% of `p_pctcpu' in 60 seconds; see CCPU_SHIFT before changing */
221 fixpt_t	ccpu = 0.95122942450071400909 * FSCALE;		/* exp(-1/20) */
222 
223 /*
224  * If `ccpu' is not equal to `exp(-1/20)' and you still want to use the
225  * faster/more-accurate formula, you'll have to estimate CCPU_SHIFT below
226  * and possibly adjust FSHIFT in "param.h" so that (FSHIFT >= CCPU_SHIFT).
227  *
228  * To estimate CCPU_SHIFT for exp(-1/20), the following formula was used:
229  *	1 - exp(-1/20) ~= 0.0487 ~= 0.0488 == 1 (fixed pt, *11* bits).
230  *
231  * If you dont want to bother with the faster/more-accurate formula, you
232  * can set CCPU_SHIFT to (FSHIFT + 1) which will use a slower/less-accurate
233  * (more general) method of calculating the %age of CPU used by a process.
234  */
235 #define	CCPU_SHIFT	11
236 
237 /*
238  * Recompute process priorities, every hz ticks.
239  */
240 /* ARGSUSED */
241 void
242 schedcpu(void *arg)
243 {
244 	fixpt_t loadfac = loadfactor(averunnable.ldavg[0]);
245 	struct proc *p;
246 	int s, s1;
247 	unsigned int newcpu;
248 	int clkhz;
249 
250 	proclist_lock_read();
251 	for (p = allproc.lh_first; p != 0; p = p->p_list.le_next) {
252 		/*
253 		 * Increment time in/out of memory and sleep time
254 		 * (if sleeping).  We ignore overflow; with 16-bit int's
255 		 * (remember them?) overflow takes 45 days.
256 		 */
257 		p->p_swtime++;
258 		if (p->p_stat == SSLEEP || p->p_stat == SSTOP)
259 			p->p_slptime++;
260 		p->p_pctcpu = (p->p_pctcpu * ccpu) >> FSHIFT;
261 		/*
262 		 * If the process has slept the entire second,
263 		 * stop recalculating its priority until it wakes up.
264 		 */
265 		if (p->p_slptime > 1)
266 			continue;
267 		s = splstatclock();	/* prevent state changes */
268 		/*
269 		 * p_pctcpu is only for ps.
270 		 */
271 		clkhz = stathz != 0 ? stathz : hz;
272 #if	(FSHIFT >= CCPU_SHIFT)
273 		p->p_pctcpu += (clkhz == 100)?
274 			((fixpt_t) p->p_cpticks) << (FSHIFT - CCPU_SHIFT):
275                 	100 * (((fixpt_t) p->p_cpticks)
276 				<< (FSHIFT - CCPU_SHIFT)) / clkhz;
277 #else
278 		p->p_pctcpu += ((FSCALE - ccpu) *
279 			(p->p_cpticks * FSCALE / clkhz)) >> FSHIFT;
280 #endif
281 		p->p_cpticks = 0;
282 		newcpu = (u_int)decay_cpu(loadfac, p->p_estcpu);
283 		p->p_estcpu = newcpu;
284 		SCHED_LOCK(s1);
285 		resetpriority(p);
286 		if (p->p_priority >= PUSER) {
287 			if (p->p_stat == SRUN &&
288 			    (p->p_flag & P_INMEM) &&
289 			    (p->p_priority / PPQ) != (p->p_usrpri / PPQ)) {
290 				remrunqueue(p);
291 				p->p_priority = p->p_usrpri;
292 				setrunqueue(p);
293 			} else
294 				p->p_priority = p->p_usrpri;
295 		}
296 		SCHED_UNLOCK(s1);
297 		splx(s);
298 	}
299 	proclist_unlock_read();
300 	uvm_meter();
301 	wakeup((caddr_t)&lbolt);
302 	callout_reset(&schedcpu_ch, hz, schedcpu, NULL);
303 }
304 
305 /*
306  * Recalculate the priority of a process after it has slept for a while.
307  * For all load averages >= 1 and max p_estcpu of 255, sleeping for at
308  * least six times the loadfactor will decay p_estcpu to zero.
309  */
310 void
311 updatepri(struct proc *p)
312 {
313 	unsigned int newcpu;
314 	fixpt_t loadfac;
315 
316 	SCHED_ASSERT_LOCKED();
317 
318 	newcpu = p->p_estcpu;
319 	loadfac = loadfactor(averunnable.ldavg[0]);
320 
321 	if (p->p_slptime > 5 * loadfac)
322 		p->p_estcpu = 0;
323 	else {
324 		p->p_slptime--;	/* the first time was done in schedcpu */
325 		while (newcpu && --p->p_slptime)
326 			newcpu = (int) decay_cpu(loadfac, newcpu);
327 		p->p_estcpu = newcpu;
328 	}
329 	resetpriority(p);
330 }
331 
332 /*
333  * During autoconfiguration or after a panic, a sleep will simply
334  * lower the priority briefly to allow interrupts, then return.
335  * The priority to be used (safepri) is machine-dependent, thus this
336  * value is initialized and maintained in the machine-dependent layers.
337  * This priority will typically be 0, or the lowest priority
338  * that is safe for use on the interrupt stack; it can be made
339  * higher to block network software interrupts after panics.
340  */
341 int safepri;
342 
343 /*
344  * General sleep call.  Suspends the current process until a wakeup is
345  * performed on the specified identifier.  The process will then be made
346  * runnable with the specified priority.  Sleeps at most timo/hz seconds
347  * (0 means no timeout).  If pri includes PCATCH flag, signals are checked
348  * before and after sleeping, else signals are not checked.  Returns 0 if
349  * awakened, EWOULDBLOCK if the timeout expires.  If PCATCH is set and a
350  * signal needs to be delivered, ERESTART is returned if the current system
351  * call should be restarted if possible, and EINTR is returned if the system
352  * call should be interrupted by the signal (return EINTR).
353  *
354  * The interlock is held until the scheduler_slock is held.  The
355  * interlock will be locked before returning back to the caller
356  * unless the PNORELOCK flag is specified, in which case the
357  * interlock will always be unlocked upon return.
358  */
359 int
360 ltsleep(void *ident, int priority, const char *wmesg, int timo,
361     __volatile struct simplelock *interlock)
362 {
363 	struct proc *p = curproc;
364 	struct slpque *qp;
365 	int sig, s;
366 	int catch = priority & PCATCH;
367 	int relock = (priority & PNORELOCK) == 0;
368 
369 	/*
370 	 * XXXSMP
371 	 * This is probably bogus.  Figure out what the right
372 	 * thing to do here really is.
373 	 * Note that not sleeping if ltsleep is called with curproc == NULL
374 	 * in the shutdown case is disgusting but partly necessary given
375 	 * how shutdown (barely) works.
376 	 */
377 	if (cold || (doing_shutdown && (panicstr || (p == NULL)))) {
378 		/*
379 		 * After a panic, or during autoconfiguration,
380 		 * just give interrupts a chance, then just return;
381 		 * don't run any other procs or panic below,
382 		 * in case this is the idle process and already asleep.
383 		 */
384 		s = splhigh();
385 		splx(safepri);
386 		splx(s);
387 		if (interlock != NULL && relock == 0)
388 			simple_unlock(interlock);
389 		return (0);
390 	}
391 
392 
393 #ifdef KTRACE
394 	if (KTRPOINT(p, KTR_CSW))
395 		ktrcsw(p, 1, 0);
396 #endif
397 
398 	SCHED_LOCK(s);
399 
400 #ifdef DIAGNOSTIC
401 	if (ident == NULL)
402 		panic("ltsleep: ident == NULL");
403 	if (p->p_stat != SONPROC)
404 		panic("ltsleep: p_stat %d != SONPROC", p->p_stat);
405 	if (p->p_back != NULL)
406 		panic("ltsleep: p_back != NULL");
407 #endif
408 
409 	p->p_wchan = ident;
410 	p->p_wmesg = wmesg;
411 	p->p_slptime = 0;
412 	p->p_priority = priority & PRIMASK;
413 
414 	qp = SLPQUE(ident);
415 	if (qp->sq_head == 0)
416 		qp->sq_head = p;
417 	else
418 		*qp->sq_tailp = p;
419 	*(qp->sq_tailp = &p->p_forw) = 0;
420 
421 	if (timo)
422 		callout_reset(&p->p_tsleep_ch, timo, endtsleep, p);
423 
424 	/*
425 	 * We can now release the interlock; the scheduler_slock
426 	 * is held, so a thread can't get in to do wakeup() before
427 	 * we do the switch.
428 	 *
429 	 * XXX We leave the code block here, after inserting ourselves
430 	 * on the sleep queue, because we might want a more clever
431 	 * data structure for the sleep queues at some point.
432 	 */
433 	if (interlock != NULL)
434 		simple_unlock(interlock);
435 
436 	/*
437 	 * We put ourselves on the sleep queue and start our timeout
438 	 * before calling CURSIG, as we could stop there, and a wakeup
439 	 * or a SIGCONT (or both) could occur while we were stopped.
440 	 * A SIGCONT would cause us to be marked as SSLEEP
441 	 * without resuming us, thus we must be ready for sleep
442 	 * when CURSIG is called.  If the wakeup happens while we're
443 	 * stopped, p->p_wchan will be 0 upon return from CURSIG.
444 	 */
445 	if (catch) {
446 		p->p_flag |= P_SINTR;
447 		if ((sig = CURSIG(p)) != 0) {
448 			if (p->p_wchan != NULL)
449 				unsleep(p);
450 			p->p_stat = SONPROC;
451 			SCHED_UNLOCK(s);
452 			goto resume;
453 		}
454 		if (p->p_wchan == NULL) {
455 			catch = 0;
456 			SCHED_UNLOCK(s);
457 			goto resume;
458 		}
459 	} else
460 		sig = 0;
461 	p->p_stat = SSLEEP;
462 	p->p_stats->p_ru.ru_nvcsw++;
463 
464 	SCHED_ASSERT_LOCKED();
465 	mi_switch(p);
466 
467 #ifdef	DDB
468 	/* handy breakpoint location after process "wakes" */
469 	asm(".globl bpendtsleep ; bpendtsleep:");
470 #endif
471 
472 	SCHED_ASSERT_UNLOCKED();
473 	splx(s);
474 
475  resume:
476 	KDASSERT(p->p_cpu != NULL);
477 	KDASSERT(p->p_cpu == curcpu());
478 	p->p_cpu->ci_schedstate.spc_curpriority = p->p_usrpri;
479 
480 	p->p_flag &= ~P_SINTR;
481 	if (p->p_flag & P_TIMEOUT) {
482 		p->p_flag &= ~P_TIMEOUT;
483 		if (sig == 0) {
484 #ifdef KTRACE
485 			if (KTRPOINT(p, KTR_CSW))
486 				ktrcsw(p, 0, 0);
487 #endif
488 			if (relock && interlock != NULL)
489 				simple_lock(interlock);
490 			return (EWOULDBLOCK);
491 		}
492 	} else if (timo)
493 		callout_stop(&p->p_tsleep_ch);
494 	if (catch && (sig != 0 || (sig = CURSIG(p)) != 0)) {
495 #ifdef KTRACE
496 		if (KTRPOINT(p, KTR_CSW))
497 			ktrcsw(p, 0, 0);
498 #endif
499 		if (relock && interlock != NULL)
500 			simple_lock(interlock);
501 		if ((SIGACTION(p, sig).sa_flags & SA_RESTART) == 0)
502 			return (EINTR);
503 		return (ERESTART);
504 	}
505 #ifdef KTRACE
506 	if (KTRPOINT(p, KTR_CSW))
507 		ktrcsw(p, 0, 0);
508 #endif
509 	if (relock && interlock != NULL)
510 		simple_lock(interlock);
511 	return (0);
512 }
513 
514 /*
515  * Implement timeout for tsleep.
516  * If process hasn't been awakened (wchan non-zero),
517  * set timeout flag and undo the sleep.  If proc
518  * is stopped, just unsleep so it will remain stopped.
519  */
520 void
521 endtsleep(void *arg)
522 {
523 	struct proc *p;
524 	int s;
525 
526 	p = (struct proc *)arg;
527 
528 	SCHED_LOCK(s);
529 	if (p->p_wchan) {
530 		if (p->p_stat == SSLEEP)
531 			setrunnable(p);
532 		else
533 			unsleep(p);
534 		p->p_flag |= P_TIMEOUT;
535 	}
536 	SCHED_UNLOCK(s);
537 }
538 
539 /*
540  * Remove a process from its wait queue
541  */
542 void
543 unsleep(struct proc *p)
544 {
545 	struct slpque *qp;
546 	struct proc **hp;
547 
548 	SCHED_ASSERT_LOCKED();
549 
550 	if (p->p_wchan) {
551 		hp = &(qp = SLPQUE(p->p_wchan))->sq_head;
552 		while (*hp != p)
553 			hp = &(*hp)->p_forw;
554 		*hp = p->p_forw;
555 		if (qp->sq_tailp == &p->p_forw)
556 			qp->sq_tailp = hp;
557 		p->p_wchan = 0;
558 	}
559 }
560 
561 /*
562  * Optimized-for-wakeup() version of setrunnable().
563  */
564 __inline void
565 awaken(struct proc *p)
566 {
567 
568 	SCHED_ASSERT_LOCKED();
569 
570 	if (p->p_slptime > 1)
571 		updatepri(p);
572 	p->p_slptime = 0;
573 	p->p_stat = SRUN;
574 
575 	/*
576 	 * Since curpriority is a user priority, p->p_priority
577 	 * is always better than curpriority.
578 	 */
579 	if (p->p_flag & P_INMEM) {
580 		setrunqueue(p);
581 		KASSERT(p->p_cpu != NULL);
582 		need_resched(p->p_cpu);
583 	} else
584 		sched_wakeup(&proc0);
585 }
586 
587 #if defined(MULTIPROCESSOR) || defined(LOCKDEBUG)
588 void
589 sched_unlock_idle(void)
590 {
591 
592 	simple_unlock(&sched_lock);
593 }
594 
595 void
596 sched_lock_idle(void)
597 {
598 
599 	simple_lock(&sched_lock);
600 }
601 #endif /* MULTIPROCESSOR || LOCKDEBUG */
602 
603 /*
604  * Make all processes sleeping on the specified identifier runnable.
605  */
606 
607 void
608 wakeup(void *ident)
609 {
610 	int s;
611 
612 	SCHED_ASSERT_UNLOCKED();
613 
614 	SCHED_LOCK(s);
615 	sched_wakeup(ident);
616 	SCHED_UNLOCK(s);
617 }
618 
619 void
620 sched_wakeup(void *ident)
621 {
622 	struct slpque *qp;
623 	struct proc *p, **q;
624 
625 	SCHED_ASSERT_LOCKED();
626 
627 	qp = SLPQUE(ident);
628  restart:
629 	for (q = &qp->sq_head; (p = *q) != NULL; ) {
630 #ifdef DIAGNOSTIC
631 		if (p->p_back || (p->p_stat != SSLEEP && p->p_stat != SSTOP))
632 			panic("wakeup");
633 #endif
634 		if (p->p_wchan == ident) {
635 			p->p_wchan = 0;
636 			*q = p->p_forw;
637 			if (qp->sq_tailp == &p->p_forw)
638 				qp->sq_tailp = q;
639 			if (p->p_stat == SSLEEP) {
640 				awaken(p);
641 				goto restart;
642 			}
643 		} else
644 			q = &p->p_forw;
645 	}
646 }
647 
648 /*
649  * Make the highest priority process first in line on the specified
650  * identifier runnable.
651  */
652 void
653 wakeup_one(void *ident)
654 {
655 	struct slpque *qp;
656 	struct proc *p, **q;
657 	struct proc *best_sleepp, **best_sleepq;
658 	struct proc *best_stopp, **best_stopq;
659 	int s;
660 
661 	best_sleepp = best_stopp = NULL;
662 	best_sleepq = best_stopq = NULL;
663 
664 	SCHED_LOCK(s);
665 
666 	qp = SLPQUE(ident);
667 
668 	for (q = &qp->sq_head; (p = *q) != NULL; q = &p->p_forw) {
669 #ifdef DIAGNOSTIC
670 		if (p->p_back || (p->p_stat != SSLEEP && p->p_stat != SSTOP))
671 			panic("wakeup_one");
672 #endif
673 		if (p->p_wchan == ident) {
674 			if (p->p_stat == SSLEEP) {
675 				if (best_sleepp == NULL ||
676 				    p->p_priority < best_sleepp->p_priority) {
677 					best_sleepp = p;
678 					best_sleepq = q;
679 				}
680 			} else {
681 				if (best_stopp == NULL ||
682 				    p->p_priority < best_stopp->p_priority) {
683 					best_stopp = p;
684 					best_stopq = q;
685 				}
686 			}
687 		}
688 	}
689 
690 	/*
691 	 * Consider any SSLEEP process higher than the highest priority SSTOP
692 	 * process.
693 	 */
694 	if (best_sleepp != NULL) {
695 		p = best_sleepp;
696 		q = best_sleepq;
697 	} else {
698 		p = best_stopp;
699 		q = best_stopq;
700 	}
701 
702 	if (p != NULL) {
703 		p->p_wchan = NULL;
704 		*q = p->p_forw;
705 		if (qp->sq_tailp == &p->p_forw)
706 			qp->sq_tailp = q;
707 		if (p->p_stat == SSLEEP)
708 			awaken(p);
709 	}
710 	SCHED_UNLOCK(s);
711 }
712 
713 /*
714  * General yield call.  Puts the current process back on its run queue and
715  * performs a voluntary context switch.
716  */
717 void
718 yield(void)
719 {
720 	struct proc *p = curproc;
721 	int s;
722 
723 	SCHED_LOCK(s);
724 	p->p_priority = p->p_usrpri;
725 	p->p_stat = SRUN;
726 	setrunqueue(p);
727 	p->p_stats->p_ru.ru_nvcsw++;
728 	mi_switch(p);
729 	SCHED_ASSERT_UNLOCKED();
730 	splx(s);
731 }
732 
733 /*
734  * General preemption call.  Puts the current process back on its run queue
735  * and performs an involuntary context switch.  If a process is supplied,
736  * we switch to that process.  Otherwise, we use the normal process selection
737  * criteria.
738  */
739 void
740 preempt(struct proc *newp)
741 {
742 	struct proc *p = curproc;
743 	int s;
744 
745 	/*
746 	 * XXX Switching to a specific process is not supported yet.
747 	 */
748 	if (newp != NULL)
749 		panic("preempt: cpu_preempt not yet implemented");
750 
751 	SCHED_LOCK(s);
752 	p->p_priority = p->p_usrpri;
753 	p->p_stat = SRUN;
754 	setrunqueue(p);
755 	p->p_stats->p_ru.ru_nivcsw++;
756 	mi_switch(p);
757 	SCHED_ASSERT_UNLOCKED();
758 	splx(s);
759 }
760 
761 /*
762  * The machine independent parts of context switch.
763  * Must be called at splsched() (no higher!) and with
764  * the sched_lock held.
765  */
766 void
767 mi_switch(struct proc *p)
768 {
769 	struct schedstate_percpu *spc;
770 	struct rlimit *rlim;
771 	long s, u;
772 	struct timeval tv;
773 #if defined(MULTIPROCESSOR)
774 	int hold_count;
775 #endif
776 
777 	SCHED_ASSERT_LOCKED();
778 
779 #if defined(MULTIPROCESSOR)
780 	/*
781 	 * Release the kernel_lock, as we are about to yield the CPU.
782 	 * The scheduler lock is still held until cpu_switch()
783 	 * selects a new process and removes it from the run queue.
784 	 */
785 	if (p->p_flag & P_BIGLOCK)
786 		hold_count = spinlock_release_all(&kernel_lock);
787 #endif
788 
789 	KDASSERT(p->p_cpu != NULL);
790 	KDASSERT(p->p_cpu == curcpu());
791 
792 	spc = &p->p_cpu->ci_schedstate;
793 
794 #if defined(LOCKDEBUG) || defined(DIAGNOSTIC)
795 	spinlock_switchcheck();
796 #endif
797 #ifdef LOCKDEBUG
798 	simple_lock_switchcheck();
799 #endif
800 
801 	/*
802 	 * Compute the amount of time during which the current
803 	 * process was running, and add that to its total so far.
804 	 */
805 	microtime(&tv);
806 	u = p->p_rtime.tv_usec + (tv.tv_usec - spc->spc_runtime.tv_usec);
807 	s = p->p_rtime.tv_sec + (tv.tv_sec - spc->spc_runtime.tv_sec);
808 	if (u < 0) {
809 		u += 1000000;
810 		s--;
811 	} else if (u >= 1000000) {
812 		u -= 1000000;
813 		s++;
814 	}
815 	p->p_rtime.tv_usec = u;
816 	p->p_rtime.tv_sec = s;
817 
818 	/*
819 	 * Check if the process exceeds its cpu resource allocation.
820 	 * If over max, kill it.  In any case, if it has run for more
821 	 * than 10 minutes, reduce priority to give others a chance.
822 	 */
823 	rlim = &p->p_rlimit[RLIMIT_CPU];
824 	if (s >= rlim->rlim_cur) {
825 		if (s >= rlim->rlim_max)
826 			psignal(p, SIGKILL);
827 		else {
828 			psignal(p, SIGXCPU);
829 			if (rlim->rlim_cur < rlim->rlim_max)
830 				rlim->rlim_cur += 5;
831 		}
832 	}
833 	if (autonicetime && s > autonicetime && p->p_ucred->cr_uid &&
834 	    p->p_nice == NZERO) {
835 		p->p_nice = autoniceval + NZERO;
836 		resetpriority(p);
837 	}
838 
839 	/*
840 	 * Process is about to yield the CPU; clear the appropriate
841 	 * scheduling flags.
842 	 */
843 	spc->spc_flags &= ~SPCF_SWITCHCLEAR;
844 
845 	/*
846 	 * Pick a new current process and switch to it.  When we
847 	 * run again, we'll return back here.
848 	 */
849 	uvmexp.swtch++;
850 	cpu_switch(p);
851 
852 	/*
853 	 * Make sure that MD code released the scheduler lock before
854 	 * resuming us.
855 	 */
856 	SCHED_ASSERT_UNLOCKED();
857 
858 	/*
859 	 * We're running again; record our new start time.  We might
860 	 * be running on a new CPU now, so don't use the cache'd
861 	 * schedstate_percpu pointer.
862 	 */
863 	KDASSERT(p->p_cpu != NULL);
864 	KDASSERT(p->p_cpu == curcpu());
865 	microtime(&p->p_cpu->ci_schedstate.spc_runtime);
866 
867 #if defined(MULTIPROCESSOR)
868 	/*
869 	 * Reacquire the kernel_lock now.  We do this after we've
870 	 * released the scheduler lock to avoid deadlock, and before
871 	 * we reacquire the interlock.
872 	 */
873 	if (p->p_flag & P_BIGLOCK)
874 		spinlock_acquire_count(&kernel_lock, hold_count);
875 #endif
876 }
877 
878 /*
879  * Initialize the (doubly-linked) run queues
880  * to be empty.
881  */
882 void
883 rqinit()
884 {
885 	int i;
886 
887 	for (i = 0; i < RUNQUE_NQS; i++)
888 		sched_qs[i].ph_link = sched_qs[i].ph_rlink =
889 		    (struct proc *)&sched_qs[i];
890 }
891 
892 /*
893  * Change process state to be runnable,
894  * placing it on the run queue if it is in memory,
895  * and awakening the swapper if it isn't in memory.
896  */
897 void
898 setrunnable(struct proc *p)
899 {
900 
901 	SCHED_ASSERT_LOCKED();
902 
903 	switch (p->p_stat) {
904 	case 0:
905 	case SRUN:
906 	case SONPROC:
907 	case SZOMB:
908 	case SDEAD:
909 	default:
910 		panic("setrunnable");
911 	case SSTOP:
912 		/*
913 		 * If we're being traced (possibly because someone attached us
914 		 * while we were stopped), check for a signal from the debugger.
915 		 */
916 		if ((p->p_flag & P_TRACED) != 0 && p->p_xstat != 0) {
917 			sigaddset(&p->p_sigctx.ps_siglist, p->p_xstat);
918 			p->p_sigctx.ps_sigcheck = 1;
919 		}
920 	case SSLEEP:
921 		unsleep(p);		/* e.g. when sending signals */
922 		break;
923 
924 	case SIDL:
925 		break;
926 	}
927 	p->p_stat = SRUN;
928 	if (p->p_flag & P_INMEM)
929 		setrunqueue(p);
930 
931 	if (p->p_slptime > 1)
932 		updatepri(p);
933 	p->p_slptime = 0;
934 	if ((p->p_flag & P_INMEM) == 0)
935 		sched_wakeup((caddr_t)&proc0);
936 	else if (p->p_priority < curcpu()->ci_schedstate.spc_curpriority) {
937 		/*
938 		 * XXXSMP
939 		 * This is not exactly right.  Since p->p_cpu persists
940 		 * across a context switch, this gives us some sort
941 		 * of processor affinity.  But we need to figure out
942 		 * at what point it's better to reschedule on a different
943 		 * CPU than the last one.
944 		 */
945 		need_resched((p->p_cpu != NULL) ? p->p_cpu : curcpu());
946 	}
947 }
948 
949 /*
950  * Compute the priority of a process when running in user mode.
951  * Arrange to reschedule if the resulting priority is better
952  * than that of the current process.
953  */
954 void
955 resetpriority(struct proc *p)
956 {
957 	unsigned int newpriority;
958 
959 	SCHED_ASSERT_LOCKED();
960 
961 	newpriority = PUSER + p->p_estcpu + NICE_WEIGHT * (p->p_nice - NZERO);
962 	newpriority = min(newpriority, MAXPRI);
963 	p->p_usrpri = newpriority;
964 	if (newpriority < curcpu()->ci_schedstate.spc_curpriority) {
965 		/*
966 		 * XXXSMP
967 		 * Same applies as in setrunnable() above.
968 		 */
969 		need_resched((p->p_cpu != NULL) ? p->p_cpu : curcpu());
970 	}
971 }
972 
973 /*
974  * We adjust the priority of the current process.  The priority of a process
975  * gets worse as it accumulates CPU time.  The cpu usage estimator (p_estcpu)
976  * is increased here.  The formula for computing priorities (in kern_synch.c)
977  * will compute a different value each time p_estcpu increases. This can
978  * cause a switch, but unless the priority crosses a PPQ boundary the actual
979  * queue will not change.  The cpu usage estimator ramps up quite quickly
980  * when the process is running (linearly), and decays away exponentially, at
981  * a rate which is proportionally slower when the system is busy.  The basic
982  * principle is that the system will 90% forget that the process used a lot
983  * of CPU time in 5 * loadav seconds.  This causes the system to favor
984  * processes which haven't run much recently, and to round-robin among other
985  * processes.
986  */
987 
988 void
989 schedclock(struct proc *p)
990 {
991 	int s;
992 
993 	p->p_estcpu = ESTCPULIM(p->p_estcpu + 1);
994 
995 	SCHED_LOCK(s);
996 	resetpriority(p);
997 	SCHED_UNLOCK(s);
998 
999 	if (p->p_priority >= PUSER)
1000 		p->p_priority = p->p_usrpri;
1001 }
1002 
1003 void
1004 suspendsched()
1005 {
1006 	struct proc *p;
1007 	int s;
1008 
1009 	/*
1010 	 * Convert all non-P_SYSTEM SSLEEP or SRUN processes to SSTOP.
1011 	 */
1012 	proclist_lock_read();
1013 	SCHED_LOCK(s);
1014 	for (p = LIST_FIRST(&allproc); p != NULL; p = LIST_NEXT(p, p_list)) {
1015 		if ((p->p_flag & P_SYSTEM) != 0)
1016 			continue;
1017 		switch (p->p_stat) {
1018 		case SRUN:
1019 			if ((p->p_flag & P_INMEM) != 0)
1020 				remrunqueue(p);
1021 			/* FALLTHROUGH */
1022 		case SSLEEP:
1023 			p->p_stat = SSTOP;
1024 			break;
1025 		case SONPROC:
1026 			/*
1027 			 * XXX SMP: we need to deal with processes on
1028 			 * others CPU !
1029 			 */
1030 			break;
1031 		default:
1032 			break;
1033 		}
1034 	}
1035 	SCHED_UNLOCK(s);
1036 	proclist_unlock_read();
1037 }
1038