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 $
40 * $DragonFly: src/sys/kern/kern_synch.c,v 1.5 2003/06/21 07:54:57 dillon Exp $
43 #include "opt_ktrace.h"
45 #include <sys/param.h>
46 #include <sys/systm.h>
48 #include <sys/kernel.h>
49 #include <sys/signalvar.h>
50 #include <sys/resourcevar.h>
51 #include <sys/vmmeter.h>
52 #include <sys/sysctl.h>
55 #include <sys/ktrace.h>
57 #include <sys/xwait.h>
59 #include <machine/cpu.h>
60 #include <machine/ipl.h>
61 #include <machine/smp.h>
63 static void sched_setup __P((void *dummy));
64 SYSINIT(sched_setup, SI_SUB_KICK_SCHEDULER, SI_ORDER_FIRST, sched_setup, NULL)
69 int sched_quantum; /* Roundrobin scheduling quantum in ticks. */
71 static struct callout loadav_callout;
73 struct loadavg averunnable =
74 { {0, 0, 0}, FSCALE }; /* load average, of runnable procs */
76 * Constants for averages over 1, 5, and 15 minutes
77 * when sampling at 5 second intervals.
79 static fixpt_t cexp[3] = {
80 0.9200444146293232 * FSCALE, /* exp(-1/12) */
81 0.9834714538216174 * FSCALE, /* exp(-1/60) */
82 0.9944598480048967 * FSCALE, /* exp(-1/180) */
85 static int curpriority_cmp __P((struct proc *p));
86 static void endtsleep __P((void *));
87 static void loadav __P((void *arg));
88 static void maybe_resched __P((struct proc *chk));
89 static void roundrobin __P((void *arg));
90 static void schedcpu __P((void *arg));
91 static void updatepri __P((struct proc *p));
94 sysctl_kern_quantum(SYSCTL_HANDLER_ARGS)
98 new_val = sched_quantum * tick;
99 error = sysctl_handle_int(oidp, &new_val, 0, req);
100 if (error != 0 || req->newptr == NULL)
104 sched_quantum = new_val / tick;
105 hogticks = 2 * sched_quantum;
109 SYSCTL_PROC(_kern, OID_AUTO, quantum, CTLTYPE_INT|CTLFLAG_RW,
110 0, sizeof sched_quantum, sysctl_kern_quantum, "I", "");
113 * Compare priorities. Return:
114 * <0: priority of p < current priority
115 * 0: priority of p == current priority
116 * >0: priority of p > current priority
117 * The priorities are the normal priorities or the normal realtime priorities
118 * if p is on the same scheduler as curproc. Otherwise the process on the
119 * more realtimeish scheduler has lowest priority. As usual, a higher
120 * priority really means a lower priority.
126 int c_class, p_class;
128 c_class = RTP_PRIO_BASE(curproc->p_rtprio.type);
129 p_class = RTP_PRIO_BASE(p->p_rtprio.type);
130 if (p_class != c_class)
131 return (p_class - c_class);
132 if (p_class == RTP_PRIO_NORMAL)
133 return (((int)p->p_priority - (int)curpriority) / PPQ);
134 return ((int)p->p_rtprio.prio - (int)curproc->p_rtprio.prio);
138 * Arrange to reschedule if necessary, taking the priorities and
139 * schedulers into account.
145 struct proc *p = curproc; /* XXX */
148 * XXX idle scheduler still broken because proccess stays on idle
149 * scheduler during waits (such as when getting FS locks). If a
150 * standard process becomes runaway cpu-bound, the system can lockup
151 * due to idle-scheduler processes in wakeup never getting any cpu.
157 } else if (chk == p) {
158 /* We may need to yield if our priority has been raised. */
159 if (curpriority_cmp(chk) > 0)
161 } else if (curpriority_cmp(chk) < 0)
166 roundrobin_interval(void)
168 return (sched_quantum);
172 * Force switch among equal priority processes every 100ms.
180 struct proc *p = curproc; /* XXX */
185 forward_roundrobin();
187 if (p == 0 || RTP_PRIO_NEED_RR(p->p_rtprio.type))
191 timeout(roundrobin, NULL, sched_quantum);
195 * Constants for digital decay and forget:
196 * 90% of (p_estcpu) usage in 5 * loadav time
197 * 95% of (p_pctcpu) usage in 60 seconds (load insensitive)
198 * Note that, as ps(1) mentions, this can let percentages
199 * total over 100% (I've seen 137.9% for 3 processes).
201 * Note that schedclock() updates p_estcpu and p_cpticks asynchronously.
203 * We wish to decay away 90% of p_estcpu in (5 * loadavg) seconds.
204 * That is, the system wants to compute a value of decay such
205 * that the following for loop:
206 * for (i = 0; i < (5 * loadavg); i++)
210 * for all values of loadavg:
212 * Mathematically this loop can be expressed by saying:
213 * decay ** (5 * loadavg) ~= .1
215 * The system computes decay as:
216 * decay = (2 * loadavg) / (2 * loadavg + 1)
218 * We wish to prove that the system's computation of decay
219 * will always fulfill the equation:
220 * decay ** (5 * loadavg) ~= .1
222 * If we compute b as:
225 * decay = b / (b + 1)
227 * We now need to prove two things:
228 * 1) Given factor ** (5 * loadavg) ~= .1, prove factor == b/(b+1)
229 * 2) Given b/(b+1) ** power ~= .1, prove power == (5 * loadavg)
232 * For x close to zero, exp(x) =~ 1 + x, since
233 * exp(x) = 0! + x**1/1! + x**2/2! + ... .
234 * therefore exp(-1/b) =~ 1 - (1/b) = (b-1)/b.
235 * For x close to zero, ln(1+x) =~ x, since
236 * ln(1+x) = x - x**2/2 + x**3/3 - ... -1 < x < 1
237 * therefore ln(b/(b+1)) = ln(1 - 1/(b+1)) =~ -1/(b+1).
241 * Solve (factor)**(power) =~ .1 given power (5*loadav):
242 * solving for factor,
243 * ln(factor) =~ (-2.30/5*loadav), or
244 * factor =~ exp(-1/((5/2.30)*loadav)) =~ exp(-1/(2*loadav)) =
245 * exp(-1/b) =~ (b-1)/b =~ b/(b+1). QED
248 * Solve (factor)**(power) =~ .1 given factor == (b/(b+1)):
250 * power*ln(b/(b+1)) =~ -2.30, or
251 * power =~ 2.3 * (b + 1) = 4.6*loadav + 2.3 =~ 5*loadav. QED
253 * Actual power values for the implemented algorithm are as follows:
255 * power: 5.68 10.32 14.94 19.55
258 /* calculations for digital decay to forget 90% of usage in 5*loadav sec */
259 #define loadfactor(loadav) (2 * (loadav))
260 #define decay_cpu(loadfac, cpu) (((loadfac) * (cpu)) / ((loadfac) + FSCALE))
262 /* decay 95% of `p_pctcpu' in 60 seconds; see CCPU_SHIFT before changing */
263 static fixpt_t ccpu = 0.95122942450071400909 * FSCALE; /* exp(-1/20) */
264 SYSCTL_INT(_kern, OID_AUTO, ccpu, CTLFLAG_RD, &ccpu, 0, "");
266 /* kernel uses `FSCALE', userland (SHOULD) use kern.fscale */
267 static int fscale __unused = FSCALE;
268 SYSCTL_INT(_kern, OID_AUTO, fscale, CTLFLAG_RD, 0, FSCALE, "");
271 * If `ccpu' is not equal to `exp(-1/20)' and you still want to use the
272 * faster/more-accurate formula, you'll have to estimate CCPU_SHIFT below
273 * and possibly adjust FSHIFT in "param.h" so that (FSHIFT >= CCPU_SHIFT).
275 * To estimate CCPU_SHIFT for exp(-1/20), the following formula was used:
276 * 1 - exp(-1/20) ~= 0.0487 ~= 0.0488 == 1 (fixed pt, *11* bits).
278 * If you don't want to bother with the faster/more-accurate formula, you
279 * can set CCPU_SHIFT to (FSHIFT + 1) which will use a slower/less-accurate
280 * (more general) method of calculating the %age of CPU used by a process.
282 #define CCPU_SHIFT 11
285 * Recompute process priorities, every hz ticks.
292 register fixpt_t loadfac = loadfactor(averunnable.ldavg[0]);
293 register struct proc *p;
294 register int realstathz, s;
296 realstathz = stathz ? stathz : hz;
297 LIST_FOREACH(p, &allproc, p_list) {
299 * Increment time in/out of memory and sleep time
300 * (if sleeping). We ignore overflow; with 16-bit int's
301 * (remember them?) overflow takes 45 days.
304 if (p->p_stat == SSLEEP || p->p_stat == SSTOP)
306 p->p_pctcpu = (p->p_pctcpu * ccpu) >> FSHIFT;
308 * If the process has slept the entire second,
309 * stop recalculating its priority until it wakes up.
311 if (p->p_slptime > 1)
313 s = splhigh(); /* prevent state changes and protect run queue */
315 * p_pctcpu is only for ps.
317 #if (FSHIFT >= CCPU_SHIFT)
318 p->p_pctcpu += (realstathz == 100)?
319 ((fixpt_t) p->p_cpticks) << (FSHIFT - CCPU_SHIFT):
320 100 * (((fixpt_t) p->p_cpticks)
321 << (FSHIFT - CCPU_SHIFT)) / realstathz;
323 p->p_pctcpu += ((FSCALE - ccpu) *
324 (p->p_cpticks * FSCALE / realstathz)) >> FSHIFT;
327 p->p_estcpu = decay_cpu(loadfac, p->p_estcpu);
329 if (p->p_priority >= PUSER) {
330 if ((p != curproc) &&
332 p->p_oncpu == 0xff && /* idle */
335 (p->p_flag & P_INMEM) &&
336 (p->p_priority / PPQ) != (p->p_usrpri / PPQ)) {
338 p->p_priority = p->p_usrpri;
341 p->p_priority = p->p_usrpri;
346 wakeup((caddr_t)&lbolt);
347 timeout(schedcpu, (void *)0, hz);
351 * Recalculate the priority of a process after it has slept for a while.
352 * For all load averages >= 1 and max p_estcpu of 255, sleeping for at
353 * least six times the loadfactor will decay p_estcpu to zero.
357 register struct proc *p;
359 register unsigned int newcpu = p->p_estcpu;
360 register fixpt_t loadfac = loadfactor(averunnable.ldavg[0]);
362 if (p->p_slptime > 5 * loadfac)
365 p->p_slptime--; /* the first time was done in schedcpu */
366 while (newcpu && --p->p_slptime)
367 newcpu = decay_cpu(loadfac, newcpu);
368 p->p_estcpu = newcpu;
374 * We're only looking at 7 bits of the address; everything is
375 * aligned to 4, lots of things are aligned to greater powers
376 * of 2. Shift right by 8, i.e. drop the bottom 256 worth.
378 #define TABLESIZE 128
379 static TAILQ_HEAD(slpquehead, proc) slpque[TABLESIZE];
380 #define LOOKUP(x) (((intptr_t)(x) >> 8) & (TABLESIZE - 1))
383 * During autoconfiguration or after a panic, a sleep will simply
384 * lower the priority briefly to allow interrupts, then return.
385 * The priority to be used (safepri) is machine-dependent, thus this
386 * value is initialized and maintained in the machine-dependent layers.
387 * This priority will typically be 0, or the lowest priority
388 * that is safe for use on the interrupt stack; it can be made
389 * higher to block network software interrupts after panics.
398 sched_quantum = hz/10;
399 hogticks = 2 * sched_quantum;
400 for (i = 0; i < TABLESIZE; i++)
401 TAILQ_INIT(&slpque[i]);
405 xwait_init(struct xwait *w)
407 bzero(w, sizeof(*w));
408 TAILQ_INIT(&w->waitq);
412 * General sleep call. Suspends the current process until a wakeup is
413 * performed on the specified identifier. The process will then be made
414 * runnable with the specified priority. Sleeps at most timo/hz seconds
415 * (0 means no timeout). If pri includes PCATCH flag, signals are checked
416 * before and after sleeping, else signals are not checked. Returns 0 if
417 * awakened, EWOULDBLOCK if the timeout expires. If PCATCH is set and a
418 * signal needs to be delivered, ERESTART is returned if the current system
419 * call should be restarted if possible, and EINTR is returned if the system
420 * call should be interrupted by the signal (return EINTR).
423 tsleep(ident, priority, wmesg, timo)
428 struct proc *p = curproc;
429 int s, sig, catch = priority & PCATCH;
430 int id = LOOKUP(ident);
431 struct callout_handle thandle;
434 if (p && KTRPOINT(p, KTR_CSW))
435 ktrcsw(p->p_tracep, 1, 0);
439 if (cold || panicstr) {
441 * After a panic, or during autoconfiguration,
442 * just give interrupts a chance, then just return;
443 * don't run any other procs or panic below,
444 * in case this is the idle process and already asleep.
450 KASSERT(p != NULL, ("tsleep1"));
451 KASSERT(ident != NULL && p->p_stat == SRUN, ("tsleep"));
456 p->p_priority = priority & PRIMASK;
457 TAILQ_INSERT_TAIL(&slpque[id], p, p_procq);
459 thandle = timeout(endtsleep, (void *)p, timo);
461 * We put ourselves on the sleep queue and start our timeout
462 * before calling CURSIG, as we could stop there, and a wakeup
463 * or a SIGCONT (or both) could occur while we were stopped.
464 * A SIGCONT would cause us to be marked as SSLEEP
465 * without resuming us, thus we must be ready for sleep
466 * when CURSIG is called. If the wakeup happens while we're
467 * stopped, p->p_wchan will be 0 upon return from CURSIG.
470 p->p_flag |= P_SINTR;
471 if ((sig = CURSIG(p))) {
477 if (p->p_wchan == 0) {
484 p->p_stats->p_ru.ru_nvcsw++;
487 curpriority = p->p_usrpri;
489 p->p_flag &= ~P_SINTR;
490 if (p->p_flag & P_TIMEOUT) {
491 p->p_flag &= ~P_TIMEOUT;
494 if (KTRPOINT(p, KTR_CSW))
495 ktrcsw(p->p_tracep, 0, 0);
497 return (EWOULDBLOCK);
500 untimeout(endtsleep, (void *)p, thandle);
501 if (catch && (sig != 0 || (sig = CURSIG(p)))) {
503 if (KTRPOINT(p, KTR_CSW))
504 ktrcsw(p->p_tracep, 0, 0);
506 if (SIGISMEMBER(p->p_sigacts->ps_sigintr, sig))
511 if (KTRPOINT(p, KTR_CSW))
512 ktrcsw(p->p_tracep, 0, 0);
518 * General sleep call. Suspends the current process until a wakeup is
519 * performed on the specified xwait structure. The process will then be made
520 * runnable with the specified priority. Sleeps at most timo/hz seconds
521 * (0 means no timeout). If pri includes PCATCH flag, signals are checked
522 * before and after sleeping, else signals are not checked. Returns 0 if
523 * awakened, EWOULDBLOCK if the timeout expires. If PCATCH is set and a
524 * signal needs to be delivered, ERESTART is returned if the current system
525 * call should be restarted if possible, and EINTR is returned if the system
526 * call should be interrupted by the signal (return EINTR).
528 * If the passed generation number is different from the generation number
529 * in the xwait, return immediately.
532 xsleep(struct xwait *w, int priority, const char *wmesg, int timo, int *gen)
534 struct proc *p = curproc;
535 int s, sig, catch = priority & PCATCH;
536 struct callout_handle thandle;
539 if (p && KTRPOINT(p, KTR_CSW))
540 ktrcsw(p->p_tracep, 1, 0);
544 if (cold || panicstr) {
546 * After a panic, or during autoconfiguration,
547 * just give interrupts a chance, then just return;
548 * don't run any other procs or panic below,
549 * in case this is the idle process and already asleep.
555 KASSERT(p != NULL, ("tsleep1"));
556 KASSERT(w != NULL && p->p_stat == SRUN, ("tsleep"));
559 * If the generation number does not match we return immediately.
561 if (*gen != w->gen) {
565 if (p && KTRPOINT(p, KTR_CSW))
566 ktrcsw(p->p_tracep, 0, 0);
574 p->p_priority = priority & PRIMASK;
575 p->p_flag |= P_XSLEEP;
576 TAILQ_INSERT_TAIL(&w->waitq, p, p_procq);
578 thandle = timeout(endtsleep, (void *)p, timo);
580 * We put ourselves on the sleep queue and start our timeout
581 * before calling CURSIG, as we could stop there, and a wakeup
582 * or a SIGCONT (or both) could occur while we were stopped.
583 * A SIGCONT would cause us to be marked as SSLEEP
584 * without resuming us, thus we must be ready for sleep
585 * when CURSIG is called. If the wakeup happens while we're
586 * stopped, p->p_wchan will be 0 upon return from CURSIG.
589 p->p_flag |= P_SINTR;
590 if ((sig = CURSIG(p))) {
596 if (p->p_wchan == NULL) {
603 p->p_stats->p_ru.ru_nvcsw++;
606 curpriority = p->p_usrpri;
607 *gen = w->gen; /* update generation number */
609 p->p_flag &= ~P_SINTR;
610 if (p->p_flag & P_TIMEOUT) {
611 p->p_flag &= ~P_TIMEOUT;
614 if (KTRPOINT(p, KTR_CSW))
615 ktrcsw(p->p_tracep, 0, 0);
617 return (EWOULDBLOCK);
620 untimeout(endtsleep, (void *)p, thandle);
621 if (catch && (sig != 0 || (sig = CURSIG(p)))) {
623 if (KTRPOINT(p, KTR_CSW))
624 ktrcsw(p->p_tracep, 0, 0);
626 if (SIGISMEMBER(p->p_sigacts->ps_sigintr, sig))
631 if (KTRPOINT(p, KTR_CSW))
632 ktrcsw(p->p_tracep, 0, 0);
638 * Implement timeout for tsleep or xsleep
640 * If process hasn't been awakened (wchan non-zero),
641 * set timeout flag and undo the sleep. If proc
642 * is stopped, just unsleep so it will remain stopped.
648 register struct proc *p;
651 p = (struct proc *)arg;
654 if (p->p_stat == SSLEEP)
658 p->p_flag |= P_TIMEOUT;
664 * Remove a process from its wait queue
668 register struct proc *p;
674 if (p->p_flag & P_XSLEEP) {
675 struct xwait *w = p->p_wchan;
676 TAILQ_REMOVE(&w->waitq, p, p_procq);
677 p->p_flag &= ~P_XSLEEP;
679 TAILQ_REMOVE(&slpque[LOOKUP(p->p_wchan)], p, p_procq);
687 * Make all processes sleeping on the explicit lock structure runnable.
690 xwakeup(struct xwait *w)
697 while ((p = TAILQ_FIRST(&w->waitq)) != NULL) {
698 TAILQ_REMOVE(&w->waitq, p, p_procq);
699 KASSERT(p->p_wchan == w && (p->p_flag & P_XSLEEP),
700 ("xwakeup: wchan mismatch for %p (%p/%p) %08x", p, p->p_wchan, w, p->p_flag & P_XSLEEP));
702 p->p_flag &= ~P_XSLEEP;
703 if (p->p_stat == SSLEEP) {
704 /* OPTIMIZED EXPANSION OF setrunnable(p); */
705 if (p->p_slptime > 1)
709 if (p->p_flag & P_INMEM) {
713 p->p_flag |= P_SWAPINREQ;
714 wakeup((caddr_t)&proc0);
722 * Make all processes sleeping on the specified identifier runnable.
726 register void *ident;
728 register struct slpquehead *qp;
729 register struct proc *p;
732 int id = LOOKUP(ident);
737 for (p = TAILQ_FIRST(qp); p != NULL; p = np) {
738 np = TAILQ_NEXT(p, p_procq);
739 if (p->p_wchan == ident) {
740 TAILQ_REMOVE(qp, p, p_procq);
742 if (p->p_stat == SSLEEP) {
743 /* OPTIMIZED EXPANSION OF setrunnable(p); */
744 if (p->p_slptime > 1)
748 if (p->p_flag & P_INMEM) {
752 p->p_flag |= P_SWAPINREQ;
753 wakeup((caddr_t)&proc0);
755 /* END INLINE EXPANSION */
764 * Make a process sleeping on the specified identifier runnable.
765 * May wake more than one process if a target process is currently
770 register void *ident;
772 register struct slpquehead *qp;
773 register struct proc *p;
776 int id = LOOKUP(ident);
782 for (p = TAILQ_FIRST(qp); p != NULL; p = np) {
783 np = TAILQ_NEXT(p, p_procq);
784 if (p->p_wchan == ident) {
785 TAILQ_REMOVE(qp, p, p_procq);
787 if (p->p_stat == SSLEEP) {
788 /* OPTIMIZED EXPANSION OF setrunnable(p); */
789 if (p->p_slptime > 1)
793 if (p->p_flag & P_INMEM) {
798 p->p_flag |= P_SWAPINREQ;
799 wakeup((caddr_t)&proc0);
801 /* END INLINE EXPANSION */
810 * The machine independent parts of mi_switch().
811 * Must be called at splstatclock() or higher.
816 struct timeval new_switchtime;
817 register struct proc *p = curproc; /* XXX */
818 register struct rlimit *rlim;
822 * XXX this spl is almost unnecessary. It is partly to allow for
823 * sloppy callers that don't do it (issignal() via CURSIG() is the
824 * main offender). It is partly to work around a bug in the i386
825 * cpu_switch() (the ipl is not preserved). We ran for years
826 * without it. I think there was only a interrupt latency problem.
827 * The main caller, tsleep(), does an splx() a couple of instructions
828 * after calling here. The buggy caller, issignal(), usually calls
829 * here at spl0() and sometimes returns at splhigh(). The process
830 * then runs for a little too long at splhigh(). The ipl gets fixed
831 * when the process returns to user mode (or earlier).
833 * It would probably be better to always call here at spl0(). Callers
834 * are prepared to give up control to another process, so they must
835 * be prepared to be interrupted. The clock stuff here may not
836 * actually need splstatclock().
841 #ifdef SIMPLELOCK_DEBUG
842 if (p->p_simple_locks)
843 printf("sleep: holding simple lock\n");
846 * Compute the amount of time during which the current
847 * process was running, and add that to its total so far.
849 microuptime(&new_switchtime);
850 if (timevalcmp(&new_switchtime, &mycpu->gd_switchtime, <)) {
851 printf("microuptime() went backwards (%ld.%06ld -> %ld.%06ld)\n",
852 mycpu->gd_switchtime.tv_sec, mycpu->gd_switchtime.tv_usec,
853 new_switchtime.tv_sec, new_switchtime.tv_usec);
854 new_switchtime = mycpu->gd_switchtime;
857 (new_switchtime.tv_usec - mycpu->gd_switchtime.tv_usec) +
858 (new_switchtime.tv_sec - mycpu->gd_switchtime.tv_sec) *
863 * Check if the process exceeds its cpu resource allocation.
864 * If over max, kill it.
866 if (p->p_stat != SZOMB && p->p_limit->p_cpulimit != RLIM_INFINITY &&
867 p->p_runtime > p->p_limit->p_cpulimit) {
868 rlim = &p->p_rlimit[RLIMIT_CPU];
869 if (p->p_runtime / (rlim_t)1000000 >= rlim->rlim_max) {
870 killproc(p, "exceeded maximum CPU limit");
873 if (rlim->rlim_cur < rlim->rlim_max) {
874 /* XXX: we should make a private copy */
881 * Pick a new current process and record its start time.
882 * YYY lwkt_switch() will run the heavy weight process restoration
883 * code, which removes the target thread and process from their
884 * respective run queues to temporarily mimic 5.x behavior.
885 * YYY the userland scheduler should pick only one user process
886 * at a time to run per cpu.
889 mycpu->gd_switchtime = new_switchtime;
892 if (mycpu->gd_switchtime.tv_sec == 0)
893 microuptime(&mycpu->gd_switchtime);
894 mycpu->gd_switchticks = ticks;
900 * Change process state to be runnable,
901 * placing it on the run queue if it is in memory,
902 * and awakening the swapper if it isn't in memory.
906 register struct proc *p;
916 panic("setrunnable");
919 unsleep(p); /* e.g. when sending signals */
926 if (p->p_flag & P_INMEM)
929 if (p->p_slptime > 1)
932 if ((p->p_flag & P_INMEM) == 0) {
933 p->p_flag |= P_SWAPINREQ;
934 wakeup((caddr_t)&proc0);
941 * Compute the priority of a process when running in user mode.
942 * Arrange to reschedule if the resulting priority is better
943 * than that of the current process.
947 register struct proc *p;
949 register unsigned int newpriority;
951 if (p->p_rtprio.type == RTP_PRIO_NORMAL) {
952 newpriority = PUSER + p->p_estcpu / INVERSE_ESTCPU_WEIGHT +
953 NICE_WEIGHT * p->p_nice;
954 newpriority = min(newpriority, MAXPRI);
955 p->p_usrpri = newpriority;
961 * Compute a tenex style load average of a quantity on
962 * 1, 5 and 15 minute intervals.
973 LIST_FOREACH(p, &allproc, p_list) {
980 for (i = 0; i < 3; i++)
981 avg->ldavg[i] = (cexp[i] * avg->ldavg[i] +
982 nrun * FSCALE * (FSCALE - cexp[i])) >> FSHIFT;
985 * Schedule the next update to occur after 5 seconds, but add a
986 * random variation to avoid synchronisation with processes that
987 * run at regular intervals.
989 callout_reset(&loadav_callout, hz * 4 + (int)(random() % (hz * 2 + 1)),
999 callout_init(&loadav_callout);
1001 /* Kick off timeout driven events by calling first time. */
1008 * We adjust the priority of the current process. The priority of
1009 * a process gets worse as it accumulates CPU time. The cpu usage
1010 * estimator (p_estcpu) is increased here. resetpriority() will
1011 * compute a different priority each time p_estcpu increases by
1012 * INVERSE_ESTCPU_WEIGHT
1013 * (until MAXPRI is reached). The cpu usage estimator ramps up
1014 * quite quickly when the process is running (linearly), and decays
1015 * away exponentially, at a rate which is proportionally slower when
1016 * the system is busy. The basic principle is that the system will
1017 * 90% forget that the process used a lot of CPU time in 5 * loadav
1018 * seconds. This causes the system to favor processes which haven't
1019 * run much recently, and to round-robin among other processes.
1027 p->p_estcpu = ESTCPULIM(p->p_estcpu + 1);
1028 if ((p->p_estcpu % INVERSE_ESTCPU_WEIGHT) == 0) {
1030 if (p->p_priority >= PUSER)
1031 p->p_priority = p->p_usrpri;