|
|
1.1 root 1: /*-
2: * Copyright (c) 1982, 1986, 1990 The Regents of the University of California.
3: * Copyright (c) 1991 The Regents of the University of California.
4: * All rights reserved.
5: *
6: * Redistribution and use in source and binary forms, with or without
7: * modification, are permitted provided that the following conditions
8: * are met:
9: * 1. Redistributions of source code must retain the above copyright
10: * notice, this list of conditions and the following disclaimer.
11: * 2. Redistributions in binary form must reproduce the above copyright
12: * notice, this list of conditions and the following disclaimer in the
13: * documentation and/or other materials provided with the distribution.
14: * 3. All advertising materials mentioning features or use of this software
15: * must display the following acknowledgement:
16: * This product includes software developed by the University of
17: * California, Berkeley and its contributors.
18: * 4. Neither the name of the University nor the names of its contributors
19: * may be used to endorse or promote products derived from this software
20: * without specific prior written permission.
21: *
22: * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23: * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24: * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25: * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26: * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27: * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28: * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29: * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30: * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31: * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32: * SUCH DAMAGE.
33: *
34: * @(#)kern_synch.c 7.18 (Berkeley) 6/27/91
35: */
36:
37: #include "param.h"
38: #include "systm.h"
39: #include "proc.h"
40: #include "kernel.h"
41: #include "buf.h"
42: #include "signalvar.h"
43: #include "resourcevar.h"
44:
45: #include "machine/cpu.h"
46:
47: u_char curpri; /* usrpri of curproc */
48:
49: /*
50: * Force switch among equal priority processes every 100ms.
51: */
52: roundrobin()
53: {
54:
55: need_resched();
56: timeout(roundrobin, (caddr_t)0, hz / 10);
57: }
58:
59: /*
60: * constants for digital decay and forget
61: * 90% of (p_cpu) usage in 5*loadav time
62: * 95% of (p_pctcpu) usage in 60 seconds (load insensitive)
63: * Note that, as ps(1) mentions, this can let percentages
64: * total over 100% (I've seen 137.9% for 3 processes).
65: *
66: * Note that hardclock updates p_cpu and p_cpticks independently.
67: *
68: * We wish to decay away 90% of p_cpu in (5 * loadavg) seconds.
69: * That is, the system wants to compute a value of decay such
70: * that the following for loop:
71: * for (i = 0; i < (5 * loadavg); i++)
72: * p_cpu *= decay;
73: * will compute
74: * p_cpu *= 0.1;
75: * for all values of loadavg:
76: *
77: * Mathematically this loop can be expressed by saying:
78: * decay ** (5 * loadavg) ~= .1
79: *
80: * The system computes decay as:
81: * decay = (2 * loadavg) / (2 * loadavg + 1)
82: *
83: * We wish to prove that the system's computation of decay
84: * will always fulfill the equation:
85: * decay ** (5 * loadavg) ~= .1
86: *
87: * If we compute b as:
88: * b = 2 * loadavg
89: * then
90: * decay = b / (b + 1)
91: *
92: * We now need to prove two things:
93: * 1) Given factor ** (5 * loadavg) ~= .1, prove factor == b/(b+1)
94: * 2) Given b/(b+1) ** power ~= .1, prove power == (5 * loadavg)
95: *
96: * Facts:
97: * For x close to zero, exp(x) =~ 1 + x, since
98: * exp(x) = 0! + x**1/1! + x**2/2! + ... .
99: * therefore exp(-1/b) =~ 1 - (1/b) = (b-1)/b.
100: * For x close to zero, ln(1+x) =~ x, since
101: * ln(1+x) = x - x**2/2 + x**3/3 - ... -1 < x < 1
102: * therefore ln(b/(b+1)) = ln(1 - 1/(b+1)) =~ -1/(b+1).
103: * ln(.1) =~ -2.30
104: *
105: * Proof of (1):
106: * Solve (factor)**(power) =~ .1 given power (5*loadav):
107: * solving for factor,
108: * ln(factor) =~ (-2.30/5*loadav), or
109: * factor =~ exp(-1/((5/2.30)*loadav)) =~ exp(-1/(2*loadav)) =
110: * exp(-1/b) =~ (b-1)/b =~ b/(b+1). QED
111: *
112: * Proof of (2):
113: * Solve (factor)**(power) =~ .1 given factor == (b/(b+1)):
114: * solving for power,
115: * power*ln(b/(b+1)) =~ -2.30, or
116: * power =~ 2.3 * (b + 1) = 4.6*loadav + 2.3 =~ 5*loadav. QED
117: *
118: * Actual power values for the implemented algorithm are as follows:
119: * loadav: 1 2 3 4
120: * power: 5.68 10.32 14.94 19.55
121: */
122:
123: /* calculations for digital decay to forget 90% of usage in 5*loadav sec */
124: #define loadfactor(loadav) (2 * (loadav))
125: #define decay_cpu(loadfac, cpu) (((loadfac) * (cpu)) / ((loadfac) + FSCALE))
126:
127: /* decay 95% of `p_pctcpu' in 60 seconds; see CCPU_SHIFT before changing */
128: fixpt_t ccpu = 0.95122942450071400909 * FSCALE; /* exp(-1/20) */
129:
130: /*
131: * If `ccpu' is not equal to `exp(-1/20)' and you still want to use the
132: * faster/more-accurate formula, you'll have to estimate CCPU_SHIFT below
133: * and possibly adjust FSHIFT in "param.h" so that (FSHIFT >= CCPU_SHIFT).
134: *
135: * To estimate CCPU_SHIFT for exp(-1/20), the following formula was used:
136: * 1 - exp(-1/20) ~= 0.0487 ~= 0.0488 == 1 (fixed pt, *11* bits).
137: *
138: * If you dont want to bother with the faster/more-accurate formula, you
139: * can set CCPU_SHIFT to (FSHIFT + 1) which will use a slower/less-accurate
140: * (more general) method of calculating the %age of CPU used by a process.
141: */
142: #define CCPU_SHIFT 11
143:
144: /*
145: * Recompute process priorities, once a second
146: */
147: schedcpu()
148: {
149: register fixpt_t loadfac = loadfactor(averunnable[0]);
150: register struct proc *p;
151: register int s;
152: register unsigned int newcpu;
153:
154: wakeup((caddr_t)&lbolt);
155: for (p = allproc; p != NULL; p = p->p_nxt) {
156: /*
157: * Increment time in/out of memory and sleep time
158: * (if sleeping). We ignore overflow; with 16-bit int's
159: * (remember them?) overflow takes 45 days.
160: */
161: p->p_time++;
162: if (p->p_stat == SSLEEP || p->p_stat == SSTOP)
163: p->p_slptime++;
164: p->p_pctcpu = (p->p_pctcpu * ccpu) >> FSHIFT;
165: /*
166: * If the process has slept the entire second,
167: * stop recalculating its priority until it wakes up.
168: */
169: if (p->p_slptime > 1)
170: continue;
171: /*
172: * p_pctcpu is only for ps.
173: */
174: #if (FSHIFT >= CCPU_SHIFT)
175: p->p_pctcpu += (hz == 100)?
176: ((fixpt_t) p->p_cpticks) << (FSHIFT - CCPU_SHIFT):
177: 100 * (((fixpt_t) p->p_cpticks)
178: << (FSHIFT - CCPU_SHIFT)) / hz;
179: #else
180: p->p_pctcpu += ((FSCALE - ccpu) *
181: (p->p_cpticks * FSCALE / hz)) >> FSHIFT;
182: #endif
183: p->p_cpticks = 0;
184: newcpu = (u_int) decay_cpu(loadfac, p->p_cpu) + p->p_nice;
185: p->p_cpu = min(newcpu, UCHAR_MAX);
186: setpri(p);
187: s = splhigh(); /* prevent state changes */
188: if (p->p_pri >= PUSER) {
189: #define PPQ (128 / NQS) /* priorities per queue */
190: if ((p != curproc) &&
191: p->p_stat == SRUN &&
192: (p->p_flag & SLOAD) &&
193: (p->p_pri / PPQ) != (p->p_usrpri / PPQ)) {
194: remrq(p);
195: p->p_pri = p->p_usrpri;
196: setrq(p);
197: } else
198: p->p_pri = p->p_usrpri;
199: }
200: splx(s);
201: }
202: vmmeter();
203: if (bclnlist != NULL)
204: wakeup((caddr_t)pageproc);
205: timeout(schedcpu, (caddr_t)0, hz);
206: }
207:
208: /*
209: * Recalculate the priority of a process after it has slept for a while.
210: * For all load averages >= 1 and max p_cpu of 255, sleeping for at least
211: * six times the loadfactor will decay p_cpu to zero.
212: */
213: updatepri(p)
214: register struct proc *p;
215: {
216: register unsigned int newcpu = p->p_cpu;
217: register fixpt_t loadfac = loadfactor(averunnable[0]);
218:
219: if (p->p_slptime > 5 * loadfac)
220: p->p_cpu = 0;
221: else {
222: p->p_slptime--; /* the first time was done in schedcpu */
223: while (newcpu && --p->p_slptime)
224: newcpu = (int) decay_cpu(loadfac, newcpu);
225: p->p_cpu = min(newcpu, UCHAR_MAX);
226: }
227: setpri(p);
228: }
229:
230: #define SQSIZE 0100 /* Must be power of 2 */
231: #define HASH(x) (( (int) x >> 5) & (SQSIZE-1))
232: struct slpque {
233: struct proc *sq_head;
234: struct proc **sq_tailp;
235: } slpque[SQSIZE];
236:
237: /*
238: * During autoconfiguration or after a panic, a sleep will simply
239: * lower the priority briefly to allow interrupts, then return.
240: * The priority to be used (safepri) is machine-dependent, thus this
241: * value is initialized and maintained in the machine-dependent layers.
242: * This priority will typically be 0, or the lowest priority
243: * that is safe for use on the interrupt stack; it can be made
244: * higher to block network software interrupts after panics.
245: */
246: int safepri;
247:
248: /*
249: * General sleep call.
250: * Suspends current process until a wakeup is made on chan.
251: * The process will then be made runnable with priority pri.
252: * Sleeps at most timo/hz seconds (0 means no timeout).
253: * If pri includes PCATCH flag, signals are checked
254: * before and after sleeping, else signals are not checked.
255: * Returns 0 if awakened, EWOULDBLOCK if the timeout expires.
256: * If PCATCH is set and a signal needs to be delivered,
257: * ERESTART is returned if the current system call should be restarted
258: * if possible, and EINTR is returned if the system call should
259: * be interrupted by the signal (return EINTR).
260: */
261: tsleep(chan, pri, wmesg, timo)
262: caddr_t chan;
263: int pri;
264: char *wmesg;
265: int timo;
266: {
267: register struct proc *p = curproc;
268: register struct slpque *qp;
269: register s;
270: int sig, catch = pri & PCATCH;
271: extern int cold;
272: int endtsleep();
273:
274: s = splhigh();
275: if (cold || panicstr) {
276: /*
277: * After a panic, or during autoconfiguration,
278: * just give interrupts a chance, then just return;
279: * don't run any other procs or panic below,
280: * in case this is the idle process and already asleep.
281: */
282: splx(safepri);
283: splx(s);
284: return (0);
285: }
286: #ifdef DIAGNOSTIC
287: if (chan == 0 || p->p_stat != SRUN || p->p_rlink)
288: panic("tsleep");
289: #endif
290: p->p_wchan = chan;
291: p->p_wmesg = wmesg;
292: p->p_slptime = 0;
293: p->p_pri = pri & PRIMASK;
294: qp = &slpque[HASH(chan)];
295: if (qp->sq_head == 0)
296: qp->sq_head = p;
297: else
298: *qp->sq_tailp = p;
299: *(qp->sq_tailp = &p->p_link) = 0;
300: if (timo)
301: timeout(endtsleep, (caddr_t)p, timo);
302: /*
303: * We put ourselves on the sleep queue and start our timeout
304: * before calling CURSIG, as we could stop there, and a wakeup
305: * or a SIGCONT (or both) could occur while we were stopped.
306: * A SIGCONT would cause us to be marked as SSLEEP
307: * without resuming us, thus we must be ready for sleep
308: * when CURSIG is called. If the wakeup happens while we're
309: * stopped, p->p_wchan will be 0 upon return from CURSIG.
310: */
311: if (catch) {
312: p->p_flag |= SSINTR;
313: if (sig = CURSIG(p)) {
314: if (p->p_wchan)
315: unsleep(p);
316: p->p_stat = SRUN;
317: goto resume;
318: }
319: if (p->p_wchan == 0) {
320: catch = 0;
321: goto resume;
322: }
323: }
324: p->p_stat = SSLEEP;
325: p->p_stats->p_ru.ru_nvcsw++;
326: swtch();
327: resume:
328: curpri = p->p_usrpri;
329: splx(s);
330: p->p_flag &= ~SSINTR;
331: if (p->p_flag & STIMO) {
332: p->p_flag &= ~STIMO;
333: if (catch == 0 || sig == 0)
334: return (EWOULDBLOCK);
335: } else if (timo)
336: untimeout(endtsleep, (caddr_t)p);
337: if (catch && (sig != 0 || (sig = CURSIG(p)))) {
338: if (p->p_sigacts->ps_sigintr & sigmask(sig))
339: return (EINTR);
340: return (ERESTART);
341: }
342: return (0);
343: }
344:
345: /*
346: * Implement timeout for tsleep.
347: * If process hasn't been awakened (wchan non-zero),
348: * set timeout flag and undo the sleep. If proc
349: * is stopped, just unsleep so it will remain stopped.
350: */
351: endtsleep(p)
352: register struct proc *p;
353: {
354: int s = splhigh();
355:
356: if (p->p_wchan) {
357: if (p->p_stat == SSLEEP)
358: setrun(p);
359: else
360: unsleep(p);
361: p->p_flag |= STIMO;
362: }
363: splx(s);
364: }
365:
366: /*
367: * Short-term, non-interruptable sleep.
368: */
369: sleep(chan, pri)
370: caddr_t chan;
371: int pri;
372: {
373: register struct proc *p = curproc;
374: register struct slpque *qp;
375: register s;
376: extern int cold;
377:
378: #ifdef DIAGNOSTIC
379: if (pri > PZERO) {
380: printf("sleep called with pri %d > PZERO, wchan: %x\n",
381: pri, chan);
382: panic("old sleep");
383: }
384: #endif
385: s = splhigh();
386: if (cold || panicstr) {
387: /*
388: * After a panic, or during autoconfiguration,
389: * just give interrupts a chance, then just return;
390: * don't run any other procs or panic below,
391: * in case this is the idle process and already asleep.
392: */
393: splx(safepri);
394: splx(s);
395: return;
396: }
397: #ifdef DIAGNOSTIC
398: if (chan==0 || p->p_stat != SRUN || p->p_rlink)
399: panic("sleep");
400: #endif
401: p->p_wchan = chan;
402: p->p_wmesg = NULL;
403: p->p_slptime = 0;
404: p->p_pri = pri;
405: qp = &slpque[HASH(chan)];
406: if (qp->sq_head == 0)
407: qp->sq_head = p;
408: else
409: *qp->sq_tailp = p;
410: *(qp->sq_tailp = &p->p_link) = 0;
411: p->p_stat = SSLEEP;
412: p->p_stats->p_ru.ru_nvcsw++;
413: swtch();
414: curpri = p->p_usrpri;
415: splx(s);
416: }
417:
418: /*
419: * Remove a process from its wait queue
420: */
421: unsleep(p)
422: register struct proc *p;
423: {
424: register struct slpque *qp;
425: register struct proc **hp;
426: int s;
427:
428: s = splhigh();
429: if (p->p_wchan) {
430: hp = &(qp = &slpque[HASH(p->p_wchan)])->sq_head;
431: while (*hp != p)
432: hp = &(*hp)->p_link;
433: *hp = p->p_link;
434: if (qp->sq_tailp == &p->p_link)
435: qp->sq_tailp = hp;
436: p->p_wchan = 0;
437: }
438: splx(s);
439: }
440:
441: /*
442: * Wakeup on "chan"; set all processes
443: * sleeping on chan to run state.
444: */
445: wakeup(chan)
446: register caddr_t chan;
447: {
448: register struct slpque *qp;
449: register struct proc *p, **q;
450: int s;
451:
452: s = splhigh();
453: qp = &slpque[HASH(chan)];
454: restart:
455: for (q = &qp->sq_head; p = *q; ) {
456: #ifdef DIAGNOSTIC
457: if (p->p_rlink || p->p_stat != SSLEEP && p->p_stat != SSTOP)
458: panic("wakeup");
459: #endif
460: if (p->p_wchan == chan) {
461: p->p_wchan = 0;
462: *q = p->p_link;
463: if (qp->sq_tailp == &p->p_link)
464: qp->sq_tailp = q;
465: if (p->p_stat == SSLEEP) {
466: /* OPTIMIZED INLINE EXPANSION OF setrun(p) */
467: if (p->p_slptime > 1)
468: updatepri(p);
469: p->p_slptime = 0;
470: p->p_stat = SRUN;
471: if (p->p_flag & SLOAD)
472: setrq(p);
473: /*
474: * Since curpri is a usrpri,
475: * p->p_pri is always better than curpri.
476: */
477: if ((p->p_flag&SLOAD) == 0)
478: wakeup((caddr_t)&proc0);
479: else
480: need_resched();
481: /* END INLINE EXPANSION */
482: goto restart;
483: }
484: } else
485: q = &p->p_link;
486: }
487: splx(s);
488: }
489:
490: /*
491: * Initialize the (doubly-linked) run queues
492: * to be empty.
493: */
494: rqinit()
495: {
496: register int i;
497:
498: for (i = 0; i < NQS; i++)
499: qs[i].ph_link = qs[i].ph_rlink = (struct proc *)&qs[i];
500: }
501:
502: /*
503: * Change process state to be runnable,
504: * placing it on the run queue if it is in memory,
505: * and awakening the swapper if it isn't in memory.
506: */
507: setrun(p)
508: register struct proc *p;
509: {
510: register int s;
511:
512: s = splhigh();
513: switch (p->p_stat) {
514:
515: case 0:
516: case SWAIT:
517: case SRUN:
518: case SZOMB:
519: default:
520: panic("setrun");
521:
522: case SSTOP:
523: case SSLEEP:
524: unsleep(p); /* e.g. when sending signals */
525: break;
526:
527: case SIDL:
528: break;
529: }
530: p->p_stat = SRUN;
531: if (p->p_flag & SLOAD)
532: setrq(p);
533: splx(s);
534: if (p->p_slptime > 1)
535: updatepri(p);
536: p->p_slptime = 0;
537: if ((p->p_flag&SLOAD) == 0)
538: wakeup((caddr_t)&proc0);
539: else if (p->p_pri < curpri)
540: need_resched();
541: }
542:
543: /*
544: * Compute priority of process when running in user mode.
545: * Arrange to reschedule if the resulting priority
546: * is better than that of the current process.
547: */
548: setpri(p)
549: register struct proc *p;
550: {
551: register unsigned int newpri;
552:
553: newpri = PUSER + p->p_cpu / 4 + 2 * p->p_nice;
554: newpri = min(newpri, MAXPRI);
555: p->p_usrpri = newpri;
556: if (newpri < curpri)
557: need_resched();
558: }
This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.