sched_4bsd.c revision 1.31
1 /* $NetBSD: sched_4bsd.c,v 1.31 2017/07/08 15:15:43 maxv Exp $ */ 2 3 /* 4 * Copyright (c) 1999, 2000, 2004, 2006, 2007, 2008 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, by Charles M. Hannum, Andrew Doran, and 10 * Daniel Sieger. 11 * 12 * Redistribution and use in source and binary forms, with or without 13 * modification, are permitted provided that the following conditions 14 * are met: 15 * 1. Redistributions of source code must retain the above copyright 16 * notice, this list of conditions and the following disclaimer. 17 * 2. Redistributions in binary form must reproduce the above copyright 18 * notice, this list of conditions and the following disclaimer in the 19 * documentation and/or other materials provided with the distribution. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 22 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 23 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 24 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 25 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 26 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 27 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 28 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 29 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 30 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 31 * POSSIBILITY OF SUCH DAMAGE. 32 */ 33 34 /* 35 * Copyright (c) 1982, 1986, 1990, 1991, 1993 36 * The Regents of the University of California. All rights reserved. 37 * (c) UNIX System Laboratories, Inc. 38 * All or some portions of this file are derived from material licensed 39 * to the University of California by American Telephone and Telegraph 40 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 41 * the permission of UNIX System Laboratories, Inc. 42 * 43 * Redistribution and use in source and binary forms, with or without 44 * modification, are permitted provided that the following conditions 45 * are met: 46 * 1. Redistributions of source code must retain the above copyright 47 * notice, this list of conditions and the following disclaimer. 48 * 2. Redistributions in binary form must reproduce the above copyright 49 * notice, this list of conditions and the following disclaimer in the 50 * documentation and/or other materials provided with the distribution. 51 * 3. Neither the name of the University nor the names of its contributors 52 * may be used to endorse or promote products derived from this software 53 * without specific prior written permission. 54 * 55 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 56 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 57 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 58 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 59 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 60 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 61 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 62 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 63 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 64 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 65 * SUCH DAMAGE. 66 * 67 * @(#)kern_synch.c 8.9 (Berkeley) 5/19/95 68 */ 69 70 #include <sys/cdefs.h> 71 __KERNEL_RCSID(0, "$NetBSD: sched_4bsd.c,v 1.31 2017/07/08 15:15:43 maxv Exp $"); 72 73 #include "opt_ddb.h" 74 #include "opt_lockdebug.h" 75 #include "opt_perfctrs.h" 76 77 #include <sys/param.h> 78 #include <sys/systm.h> 79 #include <sys/callout.h> 80 #include <sys/cpu.h> 81 #include <sys/proc.h> 82 #include <sys/kernel.h> 83 #include <sys/resourcevar.h> 84 #include <sys/sched.h> 85 #include <sys/sysctl.h> 86 #include <sys/lockdebug.h> 87 #include <sys/intr.h> 88 89 static void updatepri(struct lwp *); 90 static void resetpriority(struct lwp *); 91 92 extern unsigned int sched_pstats_ticks; /* defined in kern_synch.c */ 93 94 /* Number of hardclock ticks per sched_tick() */ 95 static int rrticks __read_mostly; 96 97 /* 98 * Force switch among equal priority processes every 100ms. 99 * Called from hardclock every hz/10 == rrticks hardclock ticks. 100 * 101 * There's no need to lock anywhere in this routine, as it's 102 * CPU-local and runs at IPL_SCHED (called from clock interrupt). 103 */ 104 /* ARGSUSED */ 105 void 106 sched_tick(struct cpu_info *ci) 107 { 108 struct schedstate_percpu *spc = &ci->ci_schedstate; 109 lwp_t *l; 110 111 spc->spc_ticks = rrticks; 112 113 if (CURCPU_IDLE_P()) { 114 cpu_need_resched(ci, 0); 115 return; 116 } 117 l = ci->ci_data.cpu_onproc; 118 if (l == NULL) { 119 return; 120 } 121 switch (l->l_class) { 122 case SCHED_FIFO: 123 /* No timeslicing for FIFO jobs. */ 124 break; 125 case SCHED_RR: 126 /* Force it into mi_switch() to look for other jobs to run. */ 127 cpu_need_resched(ci, RESCHED_KPREEMPT); 128 break; 129 default: 130 if (spc->spc_flags & SPCF_SHOULDYIELD) { 131 /* 132 * Process is stuck in kernel somewhere, probably 133 * due to buggy or inefficient code. Force a 134 * kernel preemption. 135 */ 136 cpu_need_resched(ci, RESCHED_KPREEMPT); 137 } else if (spc->spc_flags & SPCF_SEENRR) { 138 /* 139 * The process has already been through a roundrobin 140 * without switching and may be hogging the CPU. 141 * Indicate that the process should yield. 142 */ 143 spc->spc_flags |= SPCF_SHOULDYIELD; 144 cpu_need_resched(ci, 0); 145 } else { 146 spc->spc_flags |= SPCF_SEENRR; 147 } 148 break; 149 } 150 } 151 152 /* 153 * Why PRIO_MAX - 2? From setpriority(2): 154 * 155 * prio is a value in the range -20 to 20. The default priority is 156 * 0; lower priorities cause more favorable scheduling. A value of 157 * 19 or 20 will schedule a process only when nothing at priority <= 158 * 0 is runnable. 159 * 160 * This gives estcpu influence over 18 priority levels, and leaves nice 161 * with 40 levels. One way to think about it is that nice has 20 levels 162 * either side of estcpu's 18. 163 */ 164 #define ESTCPU_SHIFT 11 165 #define ESTCPU_MAX ((PRIO_MAX - 2) << ESTCPU_SHIFT) 166 #define ESTCPU_ACCUM (1 << (ESTCPU_SHIFT - 1)) 167 #define ESTCPULIM(e) min((e), ESTCPU_MAX) 168 169 /* 170 * The main parameter used by this algorithm is 'l_estcpu'. It is an estimate 171 * of the recent CPU utilization of the thread. 172 * 173 * l_estcpu is: 174 * - increased each time the hardclock ticks and the thread is found to 175 * be executing, in sched_schedclock() called from hardclock() 176 * - decreased (filtered) on each sched tick, in sched_pstats_hook() 177 * If the lwp is sleeping for more than a second, we don't touch l_estcpu: it 178 * will be updated in sched_setrunnable() when the lwp wakes up, in burst mode 179 * (ie, we decrease it n times). 180 * 181 * Note that hardclock updates l_estcpu and l_cpticks independently. 182 * 183 * ----------------------------------------------------------------------------- 184 * 185 * Here we describe how l_estcpu is decreased. 186 * 187 * Constants for digital decay (filter): 188 * 90% of l_estcpu usage in (5 * loadavg) seconds 189 * 190 * We wish to decay away 90% of l_estcpu in (5 * loadavg) seconds. That is, we 191 * want to compute a value of decay such that the following loop: 192 * for (i = 0; i < (5 * loadavg); i++) 193 * l_estcpu *= decay; 194 * will result in 195 * l_estcpu *= 0.1; 196 * for all values of loadavg. 197 * 198 * Mathematically this loop can be expressed by saying: 199 * decay ** (5 * loadavg) ~= .1 200 * 201 * And finally, the corresponding value of decay we're using is: 202 * decay = (2 * loadavg) / (2 * loadavg + 1) 203 * 204 * ----------------------------------------------------------------------------- 205 * 206 * Now, let's prove that the value of decay stated above will always fulfill 207 * the equation: 208 * decay ** (5 * loadavg) ~= .1 209 * 210 * If we compute b as: 211 * b = 2 * loadavg 212 * then 213 * decay = b / (b + 1) 214 * 215 * We now need to prove two things: 216 * 1) Given [factor ** (5 * loadavg) =~ .1], prove [factor == b/(b+1)]. 217 * 2) Given [b/(b+1) ** power =~ .1], prove [power == (5 * loadavg)]. 218 * 219 * Facts: 220 * * For x real: exp(x) = 0! + x**1/1! + x**2/2! + ... 221 * Therefore, for x close to zero, exp(x) =~ 1 + x. 222 * In turn, for b large enough, exp(-1/b) =~ 1 - (1/b) = (b-1)/b. 223 * 224 * * For b large enough, (b-1)/b =~ b/(b+1). 225 * 226 * * For x belonging to [-1;1[, ln(1-x) = - x - x**2/2 - x**3/3 - ... 227 * Therefore ln(b/(b+1)) = ln(1 - 1/(b+1)) =~ -1/(b+1). 228 * 229 * * ln(0.1) =~ -2.30 230 * 231 * Proof of (1): 232 * factor ** (5 * loadavg) =~ 0.1 233 * => ln(factor) =~ -2.30 / (5 * loadavg) 234 * => factor =~ exp(-1 / ((5 / 2.30) * loadavg)) 235 * =~ exp(-1 / (2 * loadavg)) 236 * =~ exp(-1 / b) 237 * =~ (b - 1) / b 238 * =~ b / (b + 1) 239 * =~ (2 * loadavg) / ((2 * loadavg) + 1) 240 * 241 * Proof of (2): 242 * (b / (b + 1)) ** power =~ .1 243 * => power * ln(b / (b + 1)) =~ -2.30 244 * => power * (-1 / (b + 1)) =~ -2.30 245 * => power =~ 2.30 * (b + 1) 246 * => power =~ 4.60 * loadavg + 2.30 247 * => power =~ 5 * loadavg 248 * 249 * Conclusion: decay = (2 * loadavg) / (2 * loadavg + 1) 250 */ 251 252 /* See calculations above */ 253 #define loadfactor(loadavg) (2 * (loadavg) / ncpu) 254 255 static fixpt_t 256 decay_cpu(fixpt_t loadfac, fixpt_t estcpu) 257 { 258 259 if (estcpu == 0) { 260 return 0; 261 } 262 263 #if !defined(_LP64) 264 /* avoid 64bit arithmetics. */ 265 #define FIXPT_MAX ((fixpt_t)((UINTMAX_C(1) << sizeof(fixpt_t) * CHAR_BIT) - 1)) 266 if (__predict_true(loadfac <= FIXPT_MAX / ESTCPU_MAX)) { 267 return estcpu * loadfac / (loadfac + FSCALE); 268 } 269 #endif 270 271 return (uint64_t)estcpu * loadfac / (loadfac + FSCALE); 272 } 273 274 static fixpt_t 275 decay_cpu_batch(fixpt_t loadfac, fixpt_t estcpu, unsigned int n) 276 { 277 278 /* 279 * For all load averages >= 1 and max l_estcpu of (255 << ESTCPU_SHIFT), 280 * if we slept for at least seven times the loadfactor, we will decay 281 * l_estcpu to less than (1 << ESTCPU_SHIFT), and therefore we can 282 * return zero directly. 283 * 284 * Note that our ESTCPU_MAX is actually much smaller than 285 * (255 << ESTCPU_SHIFT). 286 */ 287 if ((n << FSHIFT) >= 7 * loadfac) { 288 return 0; 289 } 290 291 while (estcpu != 0 && n > 1) { 292 estcpu = decay_cpu(loadfac, estcpu); 293 n--; 294 } 295 296 return estcpu; 297 } 298 299 /* 300 * sched_pstats_hook: 301 * 302 * Periodically called from sched_pstats(); used to recalculate priorities. 303 */ 304 void 305 sched_pstats_hook(struct lwp *l, int batch) 306 { 307 fixpt_t loadfac; 308 309 /* 310 * If the LWP has slept an entire second, stop recalculating 311 * its priority until it wakes up. 312 */ 313 KASSERT(lwp_locked(l, NULL)); 314 if (l->l_stat == LSSLEEP || l->l_stat == LSSTOP || 315 l->l_stat == LSSUSPENDED) { 316 if (l->l_slptime > 1) { 317 return; 318 } 319 } 320 loadfac = 2 * (averunnable.ldavg[0]); /* XXX: should be loadfactor? */ 321 l->l_estcpu = decay_cpu(loadfac, l->l_estcpu); 322 resetpriority(l); 323 } 324 325 /* 326 * Recalculate the priority of an LWP after it has slept for a while. 327 */ 328 static void 329 updatepri(struct lwp *l) 330 { 331 fixpt_t loadfac; 332 333 KASSERT(lwp_locked(l, NULL)); 334 KASSERT(l->l_slptime > 1); 335 336 loadfac = loadfactor(averunnable.ldavg[0]); 337 338 l->l_slptime--; /* the first time was done in sched_pstats */ 339 l->l_estcpu = decay_cpu_batch(loadfac, l->l_estcpu, l->l_slptime); 340 resetpriority(l); 341 } 342 343 void 344 sched_rqinit(void) 345 { 346 347 } 348 349 void 350 sched_setrunnable(struct lwp *l) 351 { 352 353 if (l->l_slptime > 1) 354 updatepri(l); 355 } 356 357 void 358 sched_nice(struct proc *p, int n) 359 { 360 struct lwp *l; 361 362 KASSERT(mutex_owned(p->p_lock)); 363 364 p->p_nice = n; 365 LIST_FOREACH(l, &p->p_lwps, l_sibling) { 366 lwp_lock(l); 367 resetpriority(l); 368 lwp_unlock(l); 369 } 370 } 371 372 /* 373 * Recompute the priority of an LWP. Arrange to reschedule if 374 * the resulting priority is better than that of the current LWP. 375 */ 376 static void 377 resetpriority(struct lwp *l) 378 { 379 pri_t pri; 380 struct proc *p = l->l_proc; 381 382 KASSERT(lwp_locked(l, NULL)); 383 384 if (l->l_class != SCHED_OTHER) 385 return; 386 387 /* See comments above ESTCPU_SHIFT definition. */ 388 pri = (PRI_KERNEL - 1) - (l->l_estcpu >> ESTCPU_SHIFT) - p->p_nice; 389 pri = imax(pri, 0); 390 if (pri != l->l_priority) 391 lwp_changepri(l, pri); 392 } 393 394 /* 395 * We adjust the priority of the current LWP. The priority of a LWP 396 * gets worse as it accumulates CPU time. The CPU usage estimator (l_estcpu) 397 * is increased here. The formula for computing priorities will compute a 398 * different value each time l_estcpu increases. This can cause a switch, 399 * but unless the priority crosses a PPQ boundary the actual queue will not 400 * change. The CPU usage estimator ramps up quite quickly when the process 401 * is running (linearly), and decays away exponentially, at a rate which is 402 * proportionally slower when the system is busy. The basic principle is 403 * that the system will 90% forget that the process used a lot of CPU time 404 * in (5 * loadavg) seconds. This causes the system to favor processes which 405 * haven't run much recently, and to round-robin among other processes. 406 */ 407 void 408 sched_schedclock(struct lwp *l) 409 { 410 411 if (l->l_class != SCHED_OTHER) 412 return; 413 414 KASSERT(!CURCPU_IDLE_P()); 415 l->l_estcpu = ESTCPULIM(l->l_estcpu + ESTCPU_ACCUM); 416 lwp_lock(l); 417 resetpriority(l); 418 lwp_unlock(l); 419 } 420 421 /* 422 * sched_proc_fork: 423 * 424 * Inherit the parent's scheduler history. 425 */ 426 void 427 sched_proc_fork(struct proc *parent, struct proc *child) 428 { 429 lwp_t *pl; 430 431 KASSERT(mutex_owned(parent->p_lock)); 432 433 pl = LIST_FIRST(&parent->p_lwps); 434 child->p_estcpu_inherited = pl->l_estcpu; 435 child->p_forktime = sched_pstats_ticks; 436 } 437 438 /* 439 * sched_proc_exit: 440 * 441 * Chargeback parents for the sins of their children. 442 */ 443 void 444 sched_proc_exit(struct proc *parent, struct proc *child) 445 { 446 fixpt_t loadfac = loadfactor(averunnable.ldavg[0]); 447 fixpt_t estcpu; 448 lwp_t *pl, *cl; 449 450 /* XXX Only if parent != init?? */ 451 452 mutex_enter(parent->p_lock); 453 pl = LIST_FIRST(&parent->p_lwps); 454 cl = LIST_FIRST(&child->p_lwps); 455 estcpu = decay_cpu_batch(loadfac, child->p_estcpu_inherited, 456 sched_pstats_ticks - child->p_forktime); 457 if (cl->l_estcpu > estcpu) { 458 lwp_lock(pl); 459 pl->l_estcpu = ESTCPULIM(pl->l_estcpu + cl->l_estcpu - estcpu); 460 lwp_unlock(pl); 461 } 462 mutex_exit(parent->p_lock); 463 } 464 465 void 466 sched_wakeup(struct lwp *l) 467 { 468 469 } 470 471 void 472 sched_slept(struct lwp *l) 473 { 474 475 } 476 477 void 478 sched_lwp_fork(struct lwp *l1, struct lwp *l2) 479 { 480 481 l2->l_estcpu = l1->l_estcpu; 482 } 483 484 void 485 sched_lwp_collect(struct lwp *t) 486 { 487 lwp_t *l; 488 489 /* Absorb estcpu value of collected LWP. */ 490 l = curlwp; 491 lwp_lock(l); 492 l->l_estcpu += t->l_estcpu; 493 lwp_unlock(l); 494 } 495 496 void 497 sched_oncpu(lwp_t *l) 498 { 499 500 } 501 502 void 503 sched_newts(lwp_t *l) 504 { 505 506 } 507 508 /* 509 * Sysctl nodes and initialization. 510 */ 511 512 static int 513 sysctl_sched_rtts(SYSCTLFN_ARGS) 514 { 515 struct sysctlnode node; 516 int rttsms = hztoms(rrticks); 517 518 node = *rnode; 519 node.sysctl_data = &rttsms; 520 return sysctl_lookup(SYSCTLFN_CALL(&node)); 521 } 522 523 SYSCTL_SETUP(sysctl_sched_4bsd_setup, "sysctl sched setup") 524 { 525 const struct sysctlnode *node = NULL; 526 527 sysctl_createv(clog, 0, NULL, &node, 528 CTLFLAG_PERMANENT, 529 CTLTYPE_NODE, "sched", 530 SYSCTL_DESCR("Scheduler options"), 531 NULL, 0, NULL, 0, 532 CTL_KERN, CTL_CREATE, CTL_EOL); 533 534 if (node == NULL) 535 return; 536 537 rrticks = hz / 10; 538 539 sysctl_createv(NULL, 0, &node, NULL, 540 CTLFLAG_PERMANENT, 541 CTLTYPE_STRING, "name", NULL, 542 NULL, 0, __UNCONST("4.4BSD"), 0, 543 CTL_CREATE, CTL_EOL); 544 sysctl_createv(NULL, 0, &node, NULL, 545 CTLFLAG_PERMANENT, 546 CTLTYPE_INT, "rtts", 547 SYSCTL_DESCR("Round-robin time quantum (in milliseconds)"), 548 sysctl_sched_rtts, 0, NULL, 0, 549 CTL_CREATE, CTL_EOL); 550 } 551
Indexes created Tue Sep 23 14:10:03 GMT 2025