xref: /src/sys/kern/kern_synch.c

/*	$NetBSD: kern_synch.c,v 1.68 2000/03/23 06:30:12 thorpej Exp $	*/

/*-
 * Copyright (c) 1999 The NetBSD Foundation, Inc.
 * All rights reserved.
 *
 * This code is derived from software contributed to The NetBSD Foundation
 * by Jason R. Thorpe of the Numerical Aerospace Simulation Facility,
 * NASA Ames Research Center.
 *
 * 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 NetBSD
 *	Foundation, Inc. and its contributors.
 * 4. Neither the name of The NetBSD Foundation 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 NETBSD FOUNDATION, INC. 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 FOUNDATION 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.
 */

/*-
 * 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
 */

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

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/callout.h>
#include <sys/proc.h>
#include <sys/kernel.h>
#include <sys/buf.h>
#include <sys/signalvar.h>
#include <sys/resourcevar.h>
#include <vm/vm.h>
#include <sys/sched.h>

#include <uvm/uvm_extern.h>

#ifdef KTRACE
#include <sys/ktrace.h>
#endif

#define NICE_WEIGHT 2			/* priorities per nice level */
#define	PPQ	(128 / NQS)		/* priorities per queue */

#define	ESTCPULIM(e) min((e), NICE_WEIGHT * PRIO_MAX - PPQ)

#include <machine/cpu.h>

u_char	curpriority;		/* usrpri of curproc */
int	lbolt;			/* once a second sleep address */

void roundrobin __P((void *));
void schedcpu __P((void *));
void updatepri __P((struct proc *));
void endtsleep __P((void *));

__inline void awaken __P((struct proc *));

struct callout roundrobin_ch = CALLOUT_INITIALIZER;
struct callout schedcpu_ch = CALLOUT_INITIALIZER;

/*
 * Force switch among equal priority processes every 100ms.
 */
/* ARGSUSED */
void
roundrobin(arg)
	void *arg;
{

	need_resched();
	callout_reset(&roundrobin_ch, hz / 10, 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 hardclock updates p_estcpu and p_cpticks independently.
 *
 * 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 */
fixpt_t	ccpu = 0.95122942450071400909 * FSCALE;		/* exp(-1/20) */

/*
 * 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 dont 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.
 */
/* ARGSUSED */
void
schedcpu(arg)
	void *arg;
{
	register fixpt_t loadfac = loadfactor(averunnable.ldavg[0]);
	register struct proc *p;
	register int s;
	register unsigned int newcpu;
	int clkhz;

	proclist_lock_read();
	for (p = allproc.lh_first; p != 0; p = p->p_list.le_next) {
		/*
		 * 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.
		 */
		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)
			continue;
		s = splstatclock();	/* prevent state changes */
		/*
		 * p_pctcpu is only for ps.
		 */
		clkhz = stathz != 0 ? stathz : hz;
#if	(FSHIFT >= CCPU_SHIFT)
		p->p_pctcpu += (clkhz == 100)?
			((fixpt_t) p->p_cpticks) << (FSHIFT - CCPU_SHIFT):
                	100 * (((fixpt_t) p->p_cpticks)
				<< (FSHIFT - CCPU_SHIFT)) / clkhz;
#else
		p->p_pctcpu += ((FSCALE - ccpu) *
			(p->p_cpticks * FSCALE / clkhz)) >> FSHIFT;
#endif
		p->p_cpticks = 0;
		newcpu = (u_int)decay_cpu(loadfac, p->p_estcpu);
		p->p_estcpu = newcpu;
		resetpriority(p);
		if (p->p_priority >= PUSER) {
			if ((p != curproc) &&
			    p->p_stat == SRUN &&
			    (p->p_flag & P_INMEM) &&
			    (p->p_priority / PPQ) != (p->p_usrpri / PPQ)) {
				remrunqueue(p);
				p->p_priority = p->p_usrpri;
				setrunqueue(p);
			} else
				p->p_priority = p->p_usrpri;
		}
		splx(s);
	}
	proclist_unlock_read();
	uvm_meter();
	wakeup((caddr_t)&lbolt);
	callout_reset(&schedcpu_ch, 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 = (int) 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
#define LOOKUP(x)	(((long)(x) >> 8) & (TABLESIZE - 1))
struct slpque {
	struct proc *sq_head;
	struct proc **sq_tailp;
} slpque[TABLESIZE];

/*
 * During autoconfiguration or after a panic, a sleep will simply
 * lower the priority briefly to allow interrupts, then return.
 * The priority to be used (safepri) is machine-dependent, thus this
 * value is initialized and maintained in the machine-dependent layers.
 * This priority will typically be 0, or the lowest priority
 * that is safe for use on the interrupt stack; it can be made
 * higher to block network software interrupts after panics.
 */
int safepri;

/*
 * 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).
 */
int
tsleep(ident, priority, wmesg, timo)
	void *ident;
	int priority, timo;
	const char *wmesg;
{
	register struct proc *p = curproc;
	register struct slpque *qp;
	register int s;
	int sig, catch = priority & PCATCH;
	void endtsleep __P((void *));

	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.
		 */
		s = splhigh();
		splx(safepri);
		splx(s);
		return (0);
	}

#ifdef KTRACE
	if (KTRPOINT(p, KTR_CSW))
		ktrcsw(p->p_tracep, 1, 0);
#endif
	s = splhigh();

#ifdef DIAGNOSTIC
	if (ident == NULL)
		panic("tsleep: ident == NULL");
	if (p->p_stat != SRUN)
		panic("tsleep: p_stat %d != SRUN", p->p_stat);
	if (p->p_back != NULL)
		panic("tsleep: p_back != NULL");
#endif
	p->p_wchan = ident;
	p->p_wmesg = wmesg;
	p->p_slptime = 0;
	p->p_priority = priority & PRIMASK;
	qp = &slpque[LOOKUP(ident)];
	if (qp->sq_head == 0)
		qp->sq_head = p;
	else
		*qp->sq_tailp = p;
	*(qp->sq_tailp = &p->p_forw) = 0;
	if (timo)
		callout_reset(&p->p_tsleep_ch, 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) {
		p->p_flag |= P_SINTR;
		if ((sig = CURSIG(p)) != 0) {
			if (p->p_wchan)
				unsleep(p);
			p->p_stat = SRUN;
			goto resume;
		}
		if (p->p_wchan == 0) {
			catch = 0;
			goto resume;
		}
	} else
		sig = 0;
	p->p_stat = SSLEEP;
	p->p_stats->p_ru.ru_nvcsw++;
	mi_switch();
#ifdef	DDB
	/* handy breakpoint location after process "wakes" */
	asm(".globl bpendtsleep ; bpendtsleep:");
#endif
resume:
	curpriority = p->p_usrpri;
	splx(s);
	p->p_flag &= ~P_SINTR;
	if (p->p_flag & P_TIMEOUT) {
		p->p_flag &= ~P_TIMEOUT;
		if (sig == 0) {
#ifdef KTRACE
			if (KTRPOINT(p, KTR_CSW))
				ktrcsw(p->p_tracep, 0, 0);
#endif
			return (EWOULDBLOCK);
		}
	} else if (timo)
		callout_stop(&p->p_tsleep_ch);
	if (catch && (sig != 0 || (sig = CURSIG(p)) != 0)) {
#ifdef KTRACE
		if (KTRPOINT(p, KTR_CSW))
			ktrcsw(p->p_tracep, 0, 0);
#endif
		if ((p->p_sigacts->ps_sigact[sig].sa_flags & SA_RESTART) == 0)
			return (EINTR);
		return (ERESTART);
	}
#ifdef KTRACE
	if (KTRPOINT(p, KTR_CSW))
		ktrcsw(p->p_tracep, 0, 0);
#endif
	return (0);
}

/*
 * Implement timeout for tsleep.
 * 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.
 */
void
endtsleep(arg)
	void *arg;
{
	register struct proc *p;
	int s;

	p = (struct proc *)arg;
	s = splhigh();
	if (p->p_wchan) {
		if (p->p_stat == SSLEEP)
			setrunnable(p);
		else
			unsleep(p);
		p->p_flag |= P_TIMEOUT;
	}
	splx(s);
}

/*
 * Short-term, non-interruptable sleep.
 */
void
sleep(ident, priority)
	void *ident;
	int priority;
{
	register struct proc *p = curproc;
	register struct slpque *qp;
	register int s;

#ifdef DIAGNOSTIC
	if (priority > PZERO) {
		printf("sleep called with priority %d > PZERO, wchan: %p\n",
		    priority, ident);
		panic("old sleep");
	}
#endif
	s = splhigh();
	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.
		 */
		splx(safepri);
		splx(s);
		return;
	}
#ifdef DIAGNOSTIC
	if (ident == NULL || p->p_stat != SRUN || p->p_back)
		panic("sleep");
#endif
	p->p_wchan = ident;
	p->p_wmesg = NULL;
	p->p_slptime = 0;
	p->p_priority = priority;
	qp = &slpque[LOOKUP(ident)];
	if (qp->sq_head == 0)
		qp->sq_head = p;
	else
		*qp->sq_tailp = p;
	*(qp->sq_tailp = &p->p_forw) = 0;
	p->p_stat = SSLEEP;
	p->p_stats->p_ru.ru_nvcsw++;
#ifdef KTRACE
	if (KTRPOINT(p, KTR_CSW))
		ktrcsw(p->p_tracep, 1, 0);
#endif
	mi_switch();
#ifdef	DDB
	/* handy breakpoint location after process "wakes" */
	asm(".globl bpendsleep ; bpendsleep:");
#endif
#ifdef KTRACE
	if (KTRPOINT(p, KTR_CSW))
		ktrcsw(p->p_tracep, 0, 0);
#endif
	curpriority = p->p_usrpri;
	splx(s);
}

/*
 * Remove a process from its wait queue
 */
void
unsleep(p)
	register struct proc *p;
{
	register struct slpque *qp;
	register struct proc **hp;
	int s;

	s = splhigh();
	if (p->p_wchan) {
		hp = &(qp = &slpque[LOOKUP(p->p_wchan)])->sq_head;
		while (*hp != p)
			hp = &(*hp)->p_forw;
		*hp = p->p_forw;
		if (qp->sq_tailp == &p->p_forw)
			qp->sq_tailp = hp;
		p->p_wchan = 0;
	}
	splx(s);
}

/*
 * Optimized-for-wakeup() version of setrunnable().
 */
__inline void
awaken(p)
	struct proc *p;
{

	if (p->p_slptime > 1)
		updatepri(p);
	p->p_slptime = 0;
	p->p_stat = SRUN;
	/*
	 * Since curpriority is a user priority, p->p_priority
	 * is always better than curpriority.
	 */
	if (p->p_flag & P_INMEM) {
		setrunqueue(p);
		need_resched();
	} else
		wakeup((caddr_t)&proc0);
}

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

	s = splhigh();
	qp = &slpque[LOOKUP(ident)];
restart:
	for (q = &qp->sq_head; (p = *q) != NULL; ) {
#ifdef DIAGNOSTIC
		if (p->p_back || (p->p_stat != SSLEEP && p->p_stat != SSTOP))
			panic("wakeup");
#endif
		if (p->p_wchan == ident) {
			p->p_wchan = 0;
			*q = p->p_forw;
			if (qp->sq_tailp == &p->p_forw)
				qp->sq_tailp = q;
			if (p->p_stat == SSLEEP) {
				awaken(p);
				goto restart;
			}
		} else
			q = &p->p_forw;
	}
	splx(s);
}

/*
 * Make the highest priority process first in line on the specified
 * identifier runnable.
 */
void
wakeup_one(ident)
	void *ident;
{
	struct slpque *qp;
	struct proc *p, **q;
	struct proc *best_sleepp, **best_sleepq;
	struct proc *best_stopp, **best_stopq;
	int s;

	best_sleepp = best_stopp = NULL;
	best_sleepq = best_stopq = NULL;

	s = splhigh();
	qp = &slpque[LOOKUP(ident)];
	for (q = &qp->sq_head; (p = *q) != NULL; q = &p->p_forw) {
#ifdef DIAGNOSTIC
		if (p->p_back || (p->p_stat != SSLEEP && p->p_stat != SSTOP))
			panic("wakeup_one");
#endif
		if (p->p_wchan == ident) {
			if (p->p_stat == SSLEEP) {
				if (best_sleepp == NULL ||
				    p->p_priority < best_sleepp->p_priority) {
					best_sleepp = p;
					best_sleepq = q;
				}
			} else {
				if (best_stopp == NULL ||
				    p->p_priority < best_stopp->p_priority) {
					best_stopp = p;
					best_stopq = q;
				}
			}
		}
	}

	/*
	 * Consider any SSLEEP process higher than the highest priority SSTOP
	 * process.
	 */
	if (best_sleepp != NULL) {
		p = best_sleepp;
		q = best_sleepq;
	} else {
		p = best_stopp;
		q = best_stopq;
	}

	if (p != NULL) {
		p->p_wchan = 0;
		*q = p->p_forw;
		if (qp->sq_tailp == &p->p_forw)
			qp->sq_tailp = q;
		if (p->p_stat == SSLEEP)
			awaken(p);
	}
	splx(s);
}

/*
 * The machine independent parts of mi_switch().
 * Must be called at splstatclock() or higher.
 */
void
mi_switch()
{
	register struct proc *p = curproc;	/* XXX */
	register struct rlimit *rlim;
	register long s, u;
	struct timeval tv;

#ifdef DEBUG
	if (p->p_simple_locks) {
		printf("p->p_simple_locks %d\n", p->p_simple_locks);
#ifdef LOCKDEBUG
		simple_lock_dump();
#endif
		panic("sleep: holding simple lock");
	}
#endif
	/*
	 * Compute the amount of time during which the current
	 * process was running, and add that to its total so far.
	 */
	microtime(&tv);
	u = p->p_rtime.tv_usec + (tv.tv_usec - runtime.tv_usec);
	s = p->p_rtime.tv_sec + (tv.tv_sec - runtime.tv_sec);
	if (u < 0) {
		u += 1000000;
		s--;
	} else if (u >= 1000000) {
		u -= 1000000;
		s++;
	}
	p->p_rtime.tv_usec = u;
	p->p_rtime.tv_sec = s;

	/*
	 * Check if the process exceeds its cpu resource allocation.
	 * If over max, kill it.  In any case, if it has run for more
	 * than 10 minutes, reduce priority to give others a chance.
	 */
	rlim = &p->p_rlimit[RLIMIT_CPU];
	if (s >= rlim->rlim_cur) {
		if (s >= rlim->rlim_max)
			psignal(p, SIGKILL);
		else {
			psignal(p, SIGXCPU);
			if (rlim->rlim_cur < rlim->rlim_max)
				rlim->rlim_cur += 5;
		}
	}
	if (autonicetime && s > autonicetime && p->p_ucred->cr_uid && p->p_nice == NZERO) {
		p->p_nice = autoniceval + NZERO;
		resetpriority(p);
	}

	/*
	 * Pick a new current process and record its start time.
	 */
	uvmexp.swtch++;
	cpu_switch(p);
	microtime(&runtime);
}

/*
 * Initialize the (doubly-linked) run queues
 * to be empty.
 */
void
rqinit()
{
	register int i;

	for (i = 0; i < NQS; i++)
		qs[i].ph_link = qs[i].ph_rlink = (struct proc *)&qs[i];
}

/*
 * 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;
{
	register int s;

	s = splhigh();
	switch (p->p_stat) {
	case 0:
	case SRUN:
	case SZOMB:
	case SDEAD:
	default:
		panic("setrunnable");
	case SSTOP:
		/*
		 * If we're being traced (possibly because someone attached us
		 * while we were stopped), check for a signal from the debugger.
		 */
		if ((p->p_flag & P_TRACED) != 0 && p->p_xstat != 0) {
			sigaddset(&p->p_siglist, p->p_xstat);
			p->p_sigcheck = 1;
		}
	case SSLEEP:
		unsleep(p);		/* e.g. when sending signals */
		break;

	case SIDL:
		break;
	}
	p->p_stat = SRUN;
	if (p->p_flag & P_INMEM)
		setrunqueue(p);
	splx(s);
	if (p->p_slptime > 1)
		updatepri(p);
	p->p_slptime = 0;
	if ((p->p_flag & P_INMEM) == 0)
		wakeup((caddr_t)&proc0);
	else if (p->p_priority < curpriority)
		need_resched();
}

/*
 * 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;

	newpriority = PUSER + p->p_estcpu + NICE_WEIGHT * (p->p_nice - NZERO);
	newpriority = min(newpriority, MAXPRI);
	p->p_usrpri = newpriority;
	if (newpriority < curpriority)
		need_resched();
}

/*
 * 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.  The formula for computing priorities (in kern_synch.c)
 * will compute a different value each time p_estcpu increases. This can
 * cause a switch, but unless the priority crosses a PPQ boundary the actual
 * queue will not change.  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
 * principal 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_estcpu = ESTCPULIM(p->p_estcpu + 1);
	resetpriority(p);
	if (p->p_priority >= PUSER)
		p->p_priority = p->p_usrpri;
}