2 * Copyright (c) 1982, 1986, 1990, 1991, 1993
3 * The Regents of the University of California. All rights reserved.
4 * (c) UNIX System Laboratories, Inc.
5 * All or some portions of this file are derived from material licensed
6 * to the University of California by American Telephone and Telegraph
7 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8 * the permission of UNIX System Laboratories, Inc.
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. All advertising materials mentioning features or use of this software
19 * must display the following acknowledgement:
20 * This product includes software developed by the University of
21 * California, Berkeley and its contributors.
22 * 4. Neither the name of the University nor the names of its contributors
23 * may be used to endorse or promote products derived from this software
24 * without specific prior written permission.
26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
38 * @(#)kern_synch.c 8.9 (Berkeley) 5/19/95
39 * $FreeBSD: src/sys/kern/kern_synch.c,v 1.87.2.6 2002/10/13 07:29:53 kbyanc Exp $
42 #include "opt_ktrace.h"
44 #include <sys/param.h>
45 #include <sys/systm.h>
47 #include <sys/kernel.h>
48 #include <sys/signalvar.h>
49 #include <sys/resourcevar.h>
50 #include <sys/vmmeter.h>
51 #include <sys/sysctl.h>
54 #include <sys/ktrace.h>
57 #include <machine/cpu.h>
58 #include <machine/ipl.h>
59 #include <machine/smp.h>
61 static void sched_setup __P((void *dummy));
62 SYSINIT(sched_setup, SI_SUB_KICK_SCHEDULER, SI_ORDER_FIRST, sched_setup, NULL)
67 int sched_quantum; /* Roundrobin scheduling quantum in ticks. */
69 static struct callout loadav_callout;
71 struct loadavg averunnable =
72 { {0, 0, 0}, FSCALE }; /* load average, of runnable procs */
74 * Constants for averages over 1, 5, and 15 minutes
75 * when sampling at 5 second intervals.
77 static fixpt_t cexp[3] = {
78 0.9200444146293232 * FSCALE, /* exp(-1/12) */
79 0.9834714538216174 * FSCALE, /* exp(-1/60) */
80 0.9944598480048967 * FSCALE, /* exp(-1/180) */
83 static int curpriority_cmp __P((struct proc *p));
84 static void endtsleep __P((void *));
85 static void loadav __P((void *arg));
86 static void maybe_resched __P((struct proc *chk));
87 static void roundrobin __P((void *arg));
88 static void schedcpu __P((void *arg));
89 static void updatepri __P((struct proc *p));
92 sysctl_kern_quantum(SYSCTL_HANDLER_ARGS)
96 new_val = sched_quantum * tick;
97 error = sysctl_handle_int(oidp, &new_val, 0, req);
98 if (error != 0 || req->newptr == NULL)
102 sched_quantum = new_val / tick;
103 hogticks = 2 * sched_quantum;
107 SYSCTL_PROC(_kern, OID_AUTO, quantum, CTLTYPE_INT|CTLFLAG_RW,
108 0, sizeof sched_quantum, sysctl_kern_quantum, "I", "");
111 * Compare priorities. Return:
112 * <0: priority of p < current priority
113 * 0: priority of p == current priority
114 * >0: priority of p > current priority
115 * The priorities are the normal priorities or the normal realtime priorities
116 * if p is on the same scheduler as curproc. Otherwise the process on the
117 * more realtimeish scheduler has lowest priority. As usual, a higher
118 * priority really means a lower priority.
124 int c_class, p_class;
126 c_class = RTP_PRIO_BASE(curproc->p_rtprio.type);
127 p_class = RTP_PRIO_BASE(p->p_rtprio.type);
128 if (p_class != c_class)
129 return (p_class - c_class);
130 if (p_class == RTP_PRIO_NORMAL)
131 return (((int)p->p_priority - (int)curpriority) / PPQ);
132 return ((int)p->p_rtprio.prio - (int)curproc->p_rtprio.prio);
136 * Arrange to reschedule if necessary, taking the priorities and
137 * schedulers into account.
143 struct proc *p = curproc; /* XXX */
146 * XXX idle scheduler still broken because proccess stays on idle
147 * scheduler during waits (such as when getting FS locks). If a
148 * standard process becomes runaway cpu-bound, the system can lockup
149 * due to idle-scheduler processes in wakeup never getting any cpu.
155 } else if (chk == p) {
156 /* We may need to yield if our priority has been raised. */
157 if (curpriority_cmp(chk) > 0)
159 } else if (curpriority_cmp(chk) < 0)
164 roundrobin_interval(void)
166 return (sched_quantum);
170 * Force switch among equal priority processes every 100ms.
178 struct proc *p = curproc; /* XXX */
183 forward_roundrobin();
185 if (p == 0 || RTP_PRIO_NEED_RR(p->p_rtprio.type))
189 timeout(roundrobin, NULL, sched_quantum);
193 * Constants for digital decay and forget:
194 * 90% of (p_estcpu) usage in 5 * loadav time
195 * 95% of (p_pctcpu) usage in 60 seconds (load insensitive)
196 * Note that, as ps(1) mentions, this can let percentages
197 * total over 100% (I've seen 137.9% for 3 processes).
199 * Note that schedclock() updates p_estcpu and p_cpticks asynchronously.
201 * We wish to decay away 90% of p_estcpu in (5 * loadavg) seconds.
202 * That is, the system wants to compute a value of decay such
203 * that the following for loop:
204 * for (i = 0; i < (5 * loadavg); i++)
208 * for all values of loadavg:
210 * Mathematically this loop can be expressed by saying:
211 * decay ** (5 * loadavg) ~= .1
213 * The system computes decay as:
214 * decay = (2 * loadavg) / (2 * loadavg + 1)
216 * We wish to prove that the system's computation of decay
217 * will always fulfill the equation:
218 * decay ** (5 * loadavg) ~= .1
220 * If we compute b as:
223 * decay = b / (b + 1)
225 * We now need to prove two things:
226 * 1) Given factor ** (5 * loadavg) ~= .1, prove factor == b/(b+1)
227 * 2) Given b/(b+1) ** power ~= .1, prove power == (5 * loadavg)
230 * For x close to zero, exp(x) =~ 1 + x, since
231 * exp(x) = 0! + x**1/1! + x**2/2! + ... .
232 * therefore exp(-1/b) =~ 1 - (1/b) = (b-1)/b.
233 * For x close to zero, ln(1+x) =~ x, since
234 * ln(1+x) = x - x**2/2 + x**3/3 - ... -1 < x < 1
235 * therefore ln(b/(b+1)) = ln(1 - 1/(b+1)) =~ -1/(b+1).
239 * Solve (factor)**(power) =~ .1 given power (5*loadav):
240 * solving for factor,
241 * ln(factor) =~ (-2.30/5*loadav), or
242 * factor =~ exp(-1/((5/2.30)*loadav)) =~ exp(-1/(2*loadav)) =
243 * exp(-1/b) =~ (b-1)/b =~ b/(b+1). QED
246 * Solve (factor)**(power) =~ .1 given factor == (b/(b+1)):
248 * power*ln(b/(b+1)) =~ -2.30, or
249 * power =~ 2.3 * (b + 1) = 4.6*loadav + 2.3 =~ 5*loadav. QED
251 * Actual power values for the implemented algorithm are as follows:
253 * power: 5.68 10.32 14.94 19.55
256 /* calculations for digital decay to forget 90% of usage in 5*loadav sec */
257 #define loadfactor(loadav) (2 * (loadav))
258 #define decay_cpu(loadfac, cpu) (((loadfac) * (cpu)) / ((loadfac) + FSCALE))
260 /* decay 95% of `p_pctcpu' in 60 seconds; see CCPU_SHIFT before changing */
261 static fixpt_t ccpu = 0.95122942450071400909 * FSCALE; /* exp(-1/20) */
262 SYSCTL_INT(_kern, OID_AUTO, ccpu, CTLFLAG_RD, &ccpu, 0, "");
264 /* kernel uses `FSCALE', userland (SHOULD) use kern.fscale */
265 static int fscale __unused = FSCALE;
266 SYSCTL_INT(_kern, OID_AUTO, fscale, CTLFLAG_RD, 0, FSCALE, "");
269 * If `ccpu' is not equal to `exp(-1/20)' and you still want to use the
270 * faster/more-accurate formula, you'll have to estimate CCPU_SHIFT below
271 * and possibly adjust FSHIFT in "param.h" so that (FSHIFT >= CCPU_SHIFT).
273 * To estimate CCPU_SHIFT for exp(-1/20), the following formula was used:
274 * 1 - exp(-1/20) ~= 0.0487 ~= 0.0488 == 1 (fixed pt, *11* bits).
276 * If you don't want to bother with the faster/more-accurate formula, you
277 * can set CCPU_SHIFT to (FSHIFT + 1) which will use a slower/less-accurate
278 * (more general) method of calculating the %age of CPU used by a process.
280 #define CCPU_SHIFT 11
283 * Recompute process priorities, every hz ticks.
290 register fixpt_t loadfac = loadfactor(averunnable.ldavg[0]);
291 register struct proc *p;
292 register int realstathz, s;
294 realstathz = stathz ? stathz : hz;
295 LIST_FOREACH(p, &allproc, p_list) {
297 * Increment time in/out of memory and sleep time
298 * (if sleeping). We ignore overflow; with 16-bit int's
299 * (remember them?) overflow takes 45 days.
302 if (p->p_stat == SSLEEP || p->p_stat == SSTOP)
304 p->p_pctcpu = (p->p_pctcpu * ccpu) >> FSHIFT;
306 * If the process has slept the entire second,
307 * stop recalculating its priority until it wakes up.
309 if (p->p_slptime > 1)
311 s = splhigh(); /* prevent state changes and protect run queue */
313 * p_pctcpu is only for ps.
315 #if (FSHIFT >= CCPU_SHIFT)
316 p->p_pctcpu += (realstathz == 100)?
317 ((fixpt_t) p->p_cpticks) << (FSHIFT - CCPU_SHIFT):
318 100 * (((fixpt_t) p->p_cpticks)
319 << (FSHIFT - CCPU_SHIFT)) / realstathz;
321 p->p_pctcpu += ((FSCALE - ccpu) *
322 (p->p_cpticks * FSCALE / realstathz)) >> FSHIFT;
325 p->p_estcpu = decay_cpu(loadfac, p->p_estcpu);
327 if (p->p_priority >= PUSER) {
328 if ((p != curproc) &&
330 p->p_oncpu == 0xff && /* idle */
333 (p->p_flag & P_INMEM) &&
334 (p->p_priority / PPQ) != (p->p_usrpri / PPQ)) {
336 p->p_priority = p->p_usrpri;
339 p->p_priority = p->p_usrpri;
343 wakeup((caddr_t)&lbolt);
344 timeout(schedcpu, (void *)0, hz);
348 * Recalculate the priority of a process after it has slept for a while.
349 * For all load averages >= 1 and max p_estcpu of 255, sleeping for at
350 * least six times the loadfactor will decay p_estcpu to zero.
354 register struct proc *p;
356 register unsigned int newcpu = p->p_estcpu;
357 register fixpt_t loadfac = loadfactor(averunnable.ldavg[0]);
359 if (p->p_slptime > 5 * loadfac)
362 p->p_slptime--; /* the first time was done in schedcpu */
363 while (newcpu && --p->p_slptime)
364 newcpu = decay_cpu(loadfac, newcpu);
365 p->p_estcpu = newcpu;
371 * We're only looking at 7 bits of the address; everything is
372 * aligned to 4, lots of things are aligned to greater powers
373 * of 2. Shift right by 8, i.e. drop the bottom 256 worth.
375 #define TABLESIZE 128
376 static TAILQ_HEAD(slpquehead, proc) slpque[TABLESIZE];
377 #define LOOKUP(x) (((intptr_t)(x) >> 8) & (TABLESIZE - 1))
380 * During autoconfiguration or after a panic, a sleep will simply
381 * lower the priority briefly to allow interrupts, then return.
382 * The priority to be used (safepri) is machine-dependent, thus this
383 * value is initialized and maintained in the machine-dependent layers.
384 * This priority will typically be 0, or the lowest priority
385 * that is safe for use on the interrupt stack; it can be made
386 * higher to block network software interrupts after panics.
395 sched_quantum = hz/10;
396 hogticks = 2 * sched_quantum;
397 for (i = 0; i < TABLESIZE; i++)
398 TAILQ_INIT(&slpque[i]);
402 * General sleep call. Suspends the current process until a wakeup is
403 * performed on the specified identifier. The process will then be made
404 * runnable with the specified priority. Sleeps at most timo/hz seconds
405 * (0 means no timeout). If pri includes PCATCH flag, signals are checked
406 * before and after sleeping, else signals are not checked. Returns 0 if
407 * awakened, EWOULDBLOCK if the timeout expires. If PCATCH is set and a
408 * signal needs to be delivered, ERESTART is returned if the current system
409 * call should be restarted if possible, and EINTR is returned if the system
410 * call should be interrupted by the signal (return EINTR).
413 tsleep(ident, priority, wmesg, timo)
418 struct proc *p = curproc;
419 int s, sig, catch = priority & PCATCH;
420 struct callout_handle thandle;
423 if (p && KTRPOINT(p, KTR_CSW))
424 ktrcsw(p->p_tracep, 1, 0);
427 if (cold || panicstr) {
429 * After a panic, or during autoconfiguration,
430 * just give interrupts a chance, then just return;
431 * don't run any other procs or panic below,
432 * in case this is the idle process and already asleep.
438 KASSERT(p != NULL, ("tsleep1"));
439 KASSERT(ident != NULL && p->p_stat == SRUN, ("tsleep"));
441 * Process may be sitting on a slpque if asleep() was called, remove
442 * it before re-adding.
444 if (p->p_wchan != NULL)
450 p->p_priority = priority & PRIMASK;
451 TAILQ_INSERT_TAIL(&slpque[LOOKUP(ident)], p, p_procq);
453 thandle = timeout(endtsleep, (void *)p, timo);
455 * We put ourselves on the sleep queue and start our timeout
456 * before calling CURSIG, as we could stop there, and a wakeup
457 * or a SIGCONT (or both) could occur while we were stopped.
458 * A SIGCONT would cause us to be marked as SSLEEP
459 * without resuming us, thus we must be ready for sleep
460 * when CURSIG is called. If the wakeup happens while we're
461 * stopped, p->p_wchan will be 0 upon return from CURSIG.
464 p->p_flag |= P_SINTR;
465 if ((sig = CURSIG(p))) {
471 if (p->p_wchan == 0) {
478 p->p_stats->p_ru.ru_nvcsw++;
481 curpriority = p->p_usrpri;
483 p->p_flag &= ~P_SINTR;
484 if (p->p_flag & P_TIMEOUT) {
485 p->p_flag &= ~P_TIMEOUT;
488 if (KTRPOINT(p, KTR_CSW))
489 ktrcsw(p->p_tracep, 0, 0);
491 return (EWOULDBLOCK);
494 untimeout(endtsleep, (void *)p, thandle);
495 if (catch && (sig != 0 || (sig = CURSIG(p)))) {
497 if (KTRPOINT(p, KTR_CSW))
498 ktrcsw(p->p_tracep, 0, 0);
500 if (SIGISMEMBER(p->p_sigacts->ps_sigintr, sig))
505 if (KTRPOINT(p, KTR_CSW))
506 ktrcsw(p->p_tracep, 0, 0);
512 * asleep() - async sleep call. Place process on wait queue and return
513 * immediately without blocking. The process stays runnable until await()
514 * is called. If ident is NULL, remove process from wait queue if it is still
517 * Only the most recent sleep condition is effective when making successive
518 * calls to asleep() or when calling tsleep().
520 * The timeout, if any, is not initiated until await() is called. The sleep
521 * priority, signal, and timeout is specified in the asleep() call but may be
522 * overriden in the await() call.
524 * <<<<<<<< EXPERIMENTAL, UNTESTED >>>>>>>>>>
528 asleep(void *ident, int priority, const char *wmesg, int timo)
530 struct proc *p = curproc;
534 * splhigh() while manipulating sleep structures and slpque.
536 * Remove preexisting wait condition (if any) and place process
537 * on appropriate slpque, but do not put process to sleep.
542 if (p->p_wchan != NULL)
549 p->p_asleep.as_priority = priority;
550 p->p_asleep.as_timo = timo;
551 TAILQ_INSERT_TAIL(&slpque[LOOKUP(ident)], p, p_procq);
560 * await() - wait for async condition to occur. The process blocks until
561 * wakeup() is called on the most recent asleep() address. If wakeup is called
562 * priority to await(), await() winds up being a NOP.
564 * If await() is called more then once (without an intervening asleep() call),
565 * await() is still effectively a NOP but it calls mi_switch() to give other
566 * processes some cpu before returning. The process is left runnable.
568 * <<<<<<<< EXPERIMENTAL, UNTESTED >>>>>>>>>>
572 await(int priority, int timo)
574 struct proc *p = curproc;
579 if (p->p_wchan != NULL) {
580 struct callout_handle thandle;
585 * The call to await() can override defaults specified in
586 * the original asleep().
589 priority = p->p_asleep.as_priority;
591 timo = p->p_asleep.as_timo;
598 thandle = timeout(endtsleep, (void *)p, timo);
601 catch = priority & PCATCH;
604 p->p_flag |= P_SINTR;
605 if ((sig = CURSIG(p))) {
611 if (p->p_wchan == NULL) {
617 p->p_stats->p_ru.ru_nvcsw++;
620 curpriority = p->p_usrpri;
623 p->p_flag &= ~P_SINTR;
624 if (p->p_flag & P_TIMEOUT) {
625 p->p_flag &= ~P_TIMEOUT;
628 if (KTRPOINT(p, KTR_CSW))
629 ktrcsw(p->p_tracep, 0, 0);
631 return (EWOULDBLOCK);
634 untimeout(endtsleep, (void *)p, thandle);
635 if (catch && (sig != 0 || (sig = CURSIG(p)))) {
637 if (KTRPOINT(p, KTR_CSW))
638 ktrcsw(p->p_tracep, 0, 0);
640 if (SIGISMEMBER(p->p_sigacts->ps_sigintr, sig))
645 if (KTRPOINT(p, KTR_CSW))
646 ktrcsw(p->p_tracep, 0, 0);
650 * If as_priority is 0, await() has been called without an
651 * intervening asleep(). We are still effectively a NOP,
652 * but we call mi_switch() for safety.
655 if (p->p_asleep.as_priority == 0) {
656 p->p_stats->p_ru.ru_nvcsw++;
663 * clear p_asleep.as_priority as an indication that await() has been
664 * called. If await() is called again without an intervening asleep(),
665 * await() is still effectively a NOP but the above mi_switch() code
666 * is triggered as a safety.
668 p->p_asleep.as_priority = 0;
674 * Implement timeout for tsleep or asleep()/await()
676 * If process hasn't been awakened (wchan non-zero),
677 * set timeout flag and undo the sleep. If proc
678 * is stopped, just unsleep so it will remain stopped.
684 register struct proc *p;
687 p = (struct proc *)arg;
690 if (p->p_stat == SSLEEP)
694 p->p_flag |= P_TIMEOUT;
700 * Remove a process from its wait queue
704 register struct proc *p;
710 TAILQ_REMOVE(&slpque[LOOKUP(p->p_wchan)], p, p_procq);
717 * Make all processes sleeping on the specified identifier runnable.
721 register void *ident;
723 register struct slpquehead *qp;
724 register struct proc *p;
729 qp = &slpque[LOOKUP(ident)];
731 for (p = TAILQ_FIRST(qp); p != NULL; p = np) {
732 np = TAILQ_NEXT(p, p_procq);
733 if (p->p_wchan == ident) {
734 TAILQ_REMOVE(qp, p, p_procq);
736 if (p->p_stat == SSLEEP) {
737 /* OPTIMIZED EXPANSION OF setrunnable(p); */
738 if (p->p_slptime > 1)
742 if (p->p_flag & P_INMEM) {
746 p->p_flag |= P_SWAPINREQ;
747 wakeup((caddr_t)&proc0);
749 /* END INLINE EXPANSION */
758 * Make a process sleeping on the specified identifier runnable.
759 * May wake more than one process if a target process is currently
764 register void *ident;
766 register struct slpquehead *qp;
767 register struct proc *p;
772 qp = &slpque[LOOKUP(ident)];
775 for (p = TAILQ_FIRST(qp); p != NULL; p = np) {
776 np = TAILQ_NEXT(p, p_procq);
777 if (p->p_wchan == ident) {
778 TAILQ_REMOVE(qp, p, p_procq);
780 if (p->p_stat == SSLEEP) {
781 /* OPTIMIZED EXPANSION OF setrunnable(p); */
782 if (p->p_slptime > 1)
786 if (p->p_flag & P_INMEM) {
791 p->p_flag |= P_SWAPINREQ;
792 wakeup((caddr_t)&proc0);
794 /* END INLINE EXPANSION */
803 * The machine independent parts of mi_switch().
804 * Must be called at splstatclock() or higher.
809 struct timeval new_switchtime;
810 register struct proc *p = curproc; /* XXX */
811 register struct rlimit *rlim;
815 * XXX this spl is almost unnecessary. It is partly to allow for
816 * sloppy callers that don't do it (issignal() via CURSIG() is the
817 * main offender). It is partly to work around a bug in the i386
818 * cpu_switch() (the ipl is not preserved). We ran for years
819 * without it. I think there was only a interrupt latency problem.
820 * The main caller, tsleep(), does an splx() a couple of instructions
821 * after calling here. The buggy caller, issignal(), usually calls
822 * here at spl0() and sometimes returns at splhigh(). The process
823 * then runs for a little too long at splhigh(). The ipl gets fixed
824 * when the process returns to user mode (or earlier).
826 * It would probably be better to always call here at spl0(). Callers
827 * are prepared to give up control to another process, so they must
828 * be prepared to be interrupted. The clock stuff here may not
829 * actually need splstatclock().
833 #ifdef SIMPLELOCK_DEBUG
834 if (p->p_simple_locks)
835 printf("sleep: holding simple lock\n");
838 * Compute the amount of time during which the current
839 * process was running, and add that to its total so far.
841 microuptime(&new_switchtime);
842 if (timevalcmp(&new_switchtime, &switchtime, <)) {
843 printf("microuptime() went backwards (%ld.%06ld -> %ld.%06ld)\n",
844 switchtime.tv_sec, switchtime.tv_usec,
845 new_switchtime.tv_sec, new_switchtime.tv_usec);
846 new_switchtime = switchtime;
848 p->p_runtime += (new_switchtime.tv_usec - switchtime.tv_usec) +
849 (new_switchtime.tv_sec - switchtime.tv_sec) * (int64_t)1000000;
853 * Check if the process exceeds its cpu resource allocation.
854 * If over max, kill it.
856 if (p->p_stat != SZOMB && p->p_limit->p_cpulimit != RLIM_INFINITY &&
857 p->p_runtime > p->p_limit->p_cpulimit) {
858 rlim = &p->p_rlimit[RLIMIT_CPU];
859 if (p->p_runtime / (rlim_t)1000000 >= rlim->rlim_max) {
860 killproc(p, "exceeded maximum CPU limit");
863 if (rlim->rlim_cur < rlim->rlim_max) {
864 /* XXX: we should make a private copy */
871 * Pick a new current process and record its start time.
874 switchtime = new_switchtime;
876 if (switchtime.tv_sec == 0)
877 microuptime(&switchtime);
884 * Change process state to be runnable,
885 * placing it on the run queue if it is in memory,
886 * and awakening the swapper if it isn't in memory.
890 register struct proc *p;
900 panic("setrunnable");
903 unsleep(p); /* e.g. when sending signals */
910 if (p->p_flag & P_INMEM)
913 if (p->p_slptime > 1)
916 if ((p->p_flag & P_INMEM) == 0) {
917 p->p_flag |= P_SWAPINREQ;
918 wakeup((caddr_t)&proc0);
925 * Compute the priority of a process when running in user mode.
926 * Arrange to reschedule if the resulting priority is better
927 * than that of the current process.
931 register struct proc *p;
933 register unsigned int newpriority;
935 if (p->p_rtprio.type == RTP_PRIO_NORMAL) {
936 newpriority = PUSER + p->p_estcpu / INVERSE_ESTCPU_WEIGHT +
937 NICE_WEIGHT * p->p_nice;
938 newpriority = min(newpriority, MAXPRI);
939 p->p_usrpri = newpriority;
945 * Compute a tenex style load average of a quantity on
946 * 1, 5 and 15 minute intervals.
957 LIST_FOREACH(p, &allproc, p_list) {
964 for (i = 0; i < 3; i++)
965 avg->ldavg[i] = (cexp[i] * avg->ldavg[i] +
966 nrun * FSCALE * (FSCALE - cexp[i])) >> FSHIFT;
969 * Schedule the next update to occur after 5 seconds, but add a
970 * random variation to avoid synchronisation with processes that
971 * run at regular intervals.
973 callout_reset(&loadav_callout, hz * 4 + (int)(random() % (hz * 2 + 1)),
983 callout_init(&loadav_callout);
985 /* Kick off timeout driven events by calling first time. */
992 * We adjust the priority of the current process. The priority of
993 * a process gets worse as it accumulates CPU time. The cpu usage
994 * estimator (p_estcpu) is increased here. resetpriority() will
995 * compute a different priority each time p_estcpu increases by
996 * INVERSE_ESTCPU_WEIGHT
997 * (until MAXPRI is reached). The cpu usage estimator ramps up
998 * quite quickly when the process is running (linearly), and decays
999 * away exponentially, at a rate which is proportionally slower when
1000 * the system is busy. The basic principle is that the system will
1001 * 90% forget that the process used a lot of CPU time in 5 * loadav
1002 * seconds. This causes the system to favor processes which haven't
1003 * run much recently, and to round-robin among other processes.
1011 p->p_estcpu = ESTCPULIM(p->p_estcpu + 1);
1012 if ((p->p_estcpu % INVERSE_ESTCPU_WEIGHT) == 0) {
1014 if (p->p_priority >= PUSER)
1015 p->p_priority = p->p_usrpri;