sched_4bsd.c revision 1.40
1 /* $NetBSD: sched_4bsd.c,v 1.40 2019/12/01 15:34:46 ad Exp $ */ 2 3 /* 4 * Copyright (c) 1999, 2000, 2004, 2006, 2007, 2008, 2019 5 * The NetBSD Foundation, Inc. 6 * All rights reserved. 7 * 8 * This code is derived from software contributed to The NetBSD Foundation 9 * by Jason R. Thorpe of the Numerical Aerospace Simulation Facility, 10 * NASA Ames Research Center, by Charles M. Hannum, Andrew Doran, and 11 * Daniel Sieger. 12 * 13 * Redistribution and use in source and binary forms, with or without 14 * modification, are permitted provided that the following conditions 15 * are met: 16 * 1. Redistributions of source code must retain the above copyright 17 * notice, this list of conditions and the following disclaimer. 18 * 2. Redistributions in binary form must reproduce the above copyright 19 * notice, this list of conditions and the following disclaimer in the 20 * documentation and/or other materials provided with the distribution. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 23 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 24 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 25 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 26 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 27 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 28 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 29 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 30 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 31 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 32 * POSSIBILITY OF SUCH DAMAGE. 33 */ 34 35 /* 36 * Copyright (c) 1982, 1986, 1990, 1991, 1993 37 * The Regents of the University of California. All rights reserved. 38 * (c) UNIX System Laboratories, Inc. 39 * All or some portions of this file are derived from material licensed 40 * to the University of California by American Telephone and Telegraph 41 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 42 * the permission of UNIX System Laboratories, Inc. 43 * 44 * Redistribution and use in source and binary forms, with or without 45 * modification, are permitted provided that the following conditions 46 * are met: 47 * 1. Redistributions of source code must retain the above copyright 48 * notice, this list of conditions and the following disclaimer. 49 * 2. Redistributions in binary form must reproduce the above copyright 50 * notice, this list of conditions and the following disclaimer in the 51 * documentation and/or other materials provided with the distribution. 52 * 3. Neither the name of the University nor the names of its contributors 53 * may be used to endorse or promote products derived from this software 54 * without specific prior written permission. 55 * 56 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 57 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 58 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 59 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 60 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 61 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 62 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 63 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 64 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 65 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 66 * SUCH DAMAGE. 67 * 68 * @(#)kern_synch.c 8.9 (Berkeley) 5/19/95 69 */ 70 71 #include <sys/cdefs.h> 72 __KERNEL_RCSID(0, "$NetBSD: sched_4bsd.c,v 1.40 2019/12/01 15:34:46 ad Exp $"); 73 74 #include "opt_ddb.h" 75 #include "opt_lockdebug.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 #include <sys/atomic.h> 89 90 static void updatepri(struct lwp *); 91 static void resetpriority(struct lwp *); 92 93 extern unsigned int sched_pstats_ticks; /* defined in kern_synch.c */ 94 95 /* Number of hardclock ticks per sched_tick() */ 96 static int rrticks __read_mostly; 97 98 /* 99 * Force switch among equal priority processes every 100ms. 100 * Called from hardclock every hz/10 == rrticks hardclock ticks. 101 */ 102 /* ARGSUSED */ 103 void 104 sched_tick(struct cpu_info *ci) 105 { 106 struct schedstate_percpu *spc = &ci->ci_schedstate; 107 lwp_t *l; 108 109 spc->spc_ticks = rrticks; 110 111 if (CURCPU_IDLE_P()) { 112 atomic_or_uint(&ci->ci_want_resched, 113 RESCHED_IDLE | RESCHED_UPREEMPT); 114 return; 115 } 116 l = ci->ci_onproc; 117 if (l == NULL) { 118 return; 119 } 120 /* 121 * Can only be spc_lwplock or a turnstile lock at this point 122 * (if we interrupted priority inheritance trylock dance). 123 */ 124 KASSERT(l->l_mutex != spc->spc_mutex); 125 switch (l->l_class) { 126 case SCHED_FIFO: 127 /* No timeslicing for FIFO jobs. */ 128 break; 129 case SCHED_RR: 130 /* Force it into mi_switch() to look for other jobs to run. */ 131 #ifdef __HAVE_PREEMPTION 132 atomic_or_uint(&l->l_dopreempt, DOPREEMPT_ACTIVE); 133 atomic_or_uint(&ci->ci_want_resched, RESCHED_KPREEMPT); 134 #else 135 atomic_or_uint(&ci->ci_want_resched, RESCHED_UPREEMPT); 136 #endif 137 break; 138 default: 139 if (spc->spc_flags & SPCF_SHOULDYIELD) { 140 /* 141 * Process is stuck in kernel somewhere, probably 142 * due to buggy or inefficient code. Force a 143 * kernel preemption. 144 */ 145 #ifdef __HAVE_PREEMPTION 146 atomic_or_uint(&l->l_dopreempt, DOPREEMPT_ACTIVE); 147 atomic_or_uint(&ci->ci_want_resched, RESCHED_KPREEMPT); 148 #else 149 atomic_or_uint(&ci->ci_want_resched, RESCHED_UPREEMPT); 150 #endif 151 } else if (spc->spc_flags & SPCF_SEENRR) { 152 /* 153 * The process has already been through a roundrobin 154 * without switching and may be hogging the CPU. 155 * Indicate that the process should yield. 156 */ 157 spc->spc_flags |= SPCF_SHOULDYIELD; 158 atomic_or_uint(&ci->ci_want_resched, RESCHED_UPREEMPT); 159 } else { 160 spc->spc_flags |= SPCF_SEENRR; 161 } 162 break; 163 } 164 } 165 166 /* 167 * Why PRIO_MAX - 2? From setpriority(2): 168 * 169 * prio is a value in the range -20 to 20. The default priority is 170 * 0; lower priorities cause more favorable scheduling. A value of 171 * 19 or 20 will schedule a process only when nothing at priority <= 172 * 0 is runnable. 173 * 174 * This gives estcpu influence over 18 priority levels, and leaves nice 175 * with 40 levels. One way to think about it is that nice has 20 levels 176 * either side of estcpu's 18. 177 */ 178 #define ESTCPU_SHIFT 11 179 #define ESTCPU_MAX ((PRIO_MAX - 2) << ESTCPU_SHIFT) 180 #define ESTCPU_ACCUM (1 << (ESTCPU_SHIFT - 1)) 181 #define ESTCPULIM(e) uimin((e), ESTCPU_MAX) 182 183 /* 184 * The main parameter used by this algorithm is 'l_estcpu'. It is an estimate 185 * of the recent CPU utilization of the thread. 186 * 187 * l_estcpu is: 188 * - increased each time the hardclock ticks and the thread is found to 189 * be executing, in sched_schedclock() called from hardclock() 190 * - decreased (filtered) on each sched tick, in sched_pstats_hook() 191 * If the lwp is sleeping for more than a second, we don't touch l_estcpu: it 192 * will be updated in sched_setrunnable() when the lwp wakes up, in burst mode 193 * (ie, we decrease it n times). 194 * 195 * Note that hardclock updates l_estcpu and l_cpticks independently. 196 * 197 * ----------------------------------------------------------------------------- 198 * 199 * Here we describe how l_estcpu is decreased. 200 * 201 * Constants for digital decay (filter): 202 * 90% of l_estcpu usage in (5 * loadavg) seconds 203 * 204 * We wish to decay away 90% of l_estcpu in (5 * loadavg) seconds. That is, we 205 * want to compute a value of decay such that the following loop: 206 * for (i = 0; i < (5 * loadavg); i++) 207 * l_estcpu *= decay; 208 * will result in 209 * l_estcpu *= 0.1; 210 * for all values of loadavg. 211 * 212 * Mathematically this loop can be expressed by saying: 213 * decay ** (5 * loadavg) ~= .1 214 * 215 * And finally, the corresponding value of decay we're using is: 216 * decay = (2 * loadavg) / (2 * loadavg + 1) 217 * 218 * ----------------------------------------------------------------------------- 219 * 220 * Now, let's prove that the value of decay stated above will always fulfill 221 * the equation: 222 * decay ** (5 * loadavg) ~= .1 223 * 224 * If we compute b as: 225 * b = 2 * loadavg 226 * then 227 * decay = b / (b + 1) 228 * 229 * We now need to prove two things: 230 * 1) Given [factor ** (5 * loadavg) =~ .1], prove [factor == b/(b+1)]. 231 * 2) Given [b/(b+1) ** power =~ .1], prove [power == (5 * loadavg)]. 232 * 233 * Facts: 234 * * For x real: exp(x) = 0! + x**1/1! + x**2/2! + ... 235 * Therefore, for x close to zero, exp(x) =~ 1 + x. 236 * In turn, for b large enough, exp(-1/b) =~ 1 - (1/b) = (b-1)/b. 237 * 238 * * For b large enough, (b-1)/b =~ b/(b+1). 239 * 240 * * For x belonging to [-1;1[, ln(1-x) = - x - x**2/2 - x**3/3 - ... 241 * Therefore ln(b/(b+1)) = ln(1 - 1/(b+1)) =~ -1/(b+1). 242 * 243 * * ln(0.1) =~ -2.30 244 * 245 * Proof of (1): 246 * factor ** (5 * loadavg) =~ 0.1 247 * => ln(factor) =~ -2.30 / (5 * loadavg) 248 * => factor =~ exp(-1 / ((5 / 2.30) * loadavg)) 249 * =~ exp(-1 / (2 * loadavg)) 250 * =~ exp(-1 / b) 251 * =~ (b - 1) / b 252 * =~ b / (b + 1) 253 * =~ (2 * loadavg) / ((2 * loadavg) + 1) 254 * 255 * Proof of (2): 256 * (b / (b + 1)) ** power =~ .1 257 * => power * ln(b / (b + 1)) =~ -2.30 258 * => power * (-1 / (b + 1)) =~ -2.30 259 * => power =~ 2.30 * (b + 1) 260 * => power =~ 4.60 * loadavg + 2.30 261 * => power =~ 5 * loadavg 262 * 263 * Conclusion: decay = (2 * loadavg) / (2 * loadavg + 1) 264 */ 265 266 /* See calculations above */ 267 #define loadfactor(loadavg) (2 * (loadavg)) 268 269 static fixpt_t 270 decay_cpu(fixpt_t loadfac, fixpt_t estcpu) 271 { 272 273 if (estcpu == 0) { 274 return 0; 275 } 276 277 #if !defined(_LP64) 278 /* avoid 64bit arithmetics. */ 279 #define FIXPT_MAX ((fixpt_t)((UINTMAX_C(1) << sizeof(fixpt_t) * CHAR_BIT) - 1)) 280 if (__predict_true(loadfac <= FIXPT_MAX / ESTCPU_MAX)) { 281 return estcpu * loadfac / (loadfac + FSCALE); 282 } 283 #endif 284 285 return (uint64_t)estcpu * loadfac / (loadfac + FSCALE); 286 } 287 288 static fixpt_t 289 decay_cpu_batch(fixpt_t loadfac, fixpt_t estcpu, unsigned int n) 290 { 291 292 /* 293 * For all load averages >= 1 and max l_estcpu of (255 << ESTCPU_SHIFT), 294 * if we slept for at least seven times the loadfactor, we will decay 295 * l_estcpu to less than (1 << ESTCPU_SHIFT), and therefore we can 296 * return zero directly. 297 * 298 * Note that our ESTCPU_MAX is actually much smaller than 299 * (255 << ESTCPU_SHIFT). 300 */ 301 if ((n << FSHIFT) >= 7 * loadfac) { 302 return 0; 303 } 304 305 while (estcpu != 0 && n > 1) { 306 estcpu = decay_cpu(loadfac, estcpu); 307 n--; 308 } 309 310 return estcpu; 311 } 312 313 /* 314 * sched_pstats_hook: 315 * 316 * Periodically called from sched_pstats(); used to recalculate priorities. 317 */ 318 void 319 sched_pstats_hook(struct lwp *l, int batch) 320 { 321 fixpt_t loadfac; 322 323 /* 324 * If the LWP has slept an entire second, stop recalculating 325 * its priority until it wakes up. 326 */ 327 KASSERT(lwp_locked(l, NULL)); 328 if (l->l_stat == LSSLEEP || l->l_stat == LSSTOP || 329 l->l_stat == LSSUSPENDED) { 330 if (l->l_slptime > 1) { 331 return; 332 } 333 } 334 335 loadfac = loadfactor(averunnable.ldavg[0]); 336 l->l_estcpu = decay_cpu(loadfac, l->l_estcpu); 337 resetpriority(l); 338 } 339 340 /* 341 * Recalculate the priority of an LWP after it has slept for a while. 342 */ 343 static void 344 updatepri(struct lwp *l) 345 { 346 fixpt_t loadfac; 347 348 KASSERT(lwp_locked(l, NULL)); 349 KASSERT(l->l_slptime > 1); 350 351 loadfac = loadfactor(averunnable.ldavg[0]); 352 353 l->l_slptime--; /* the first time was done in sched_pstats */ 354 l->l_estcpu = decay_cpu_batch(loadfac, l->l_estcpu, l->l_slptime); 355 resetpriority(l); 356 } 357 358 void 359 sched_rqinit(void) 360 { 361 362 } 363 364 void 365 sched_setrunnable(struct lwp *l) 366 { 367 368 if (l->l_slptime > 1) 369 updatepri(l); 370 } 371 372 void 373 sched_nice(struct proc *p, int n) 374 { 375 struct lwp *l; 376 377 KASSERT(mutex_owned(p->p_lock)); 378 379 p->p_nice = n; 380 LIST_FOREACH(l, &p->p_lwps, l_sibling) { 381 lwp_lock(l); 382 resetpriority(l); 383 lwp_unlock(l); 384 } 385 } 386 387 /* 388 * Recompute the priority of an LWP. Arrange to reschedule if 389 * the resulting priority is better than that of the current LWP. 390 */ 391 static void 392 resetpriority(struct lwp *l) 393 { 394 pri_t pri; 395 struct proc *p = l->l_proc; 396 397 KASSERT(lwp_locked(l, NULL)); 398 399 if (l->l_class != SCHED_OTHER) 400 return; 401 402 /* See comments above ESTCPU_SHIFT definition. */ 403 pri = (PRI_KERNEL - 1) - (l->l_estcpu >> ESTCPU_SHIFT) - p->p_nice; 404 pri = imax(pri, 0); 405 if (pri != l->l_priority) 406 lwp_changepri(l, pri); 407 } 408 409 /* 410 * We adjust the priority of the current LWP. The priority of a LWP 411 * gets worse as it accumulates CPU time. The CPU usage estimator (l_estcpu) 412 * is increased here. The formula for computing priorities will compute a 413 * different value each time l_estcpu increases. This can cause a switch, 414 * but unless the priority crosses a PPQ boundary the actual queue will not 415 * change. The CPU usage estimator ramps up quite quickly when the process 416 * is running (linearly), and decays away exponentially, at a rate which is 417 * proportionally slower when the system is busy. The basic principle is 418 * that the system will 90% forget that the process used a lot of CPU time 419 * in (5 * loadavg) seconds. This causes the system to favor processes which 420 * haven't run much recently, and to round-robin among other processes. 421 */ 422 void 423 sched_schedclock(struct lwp *l) 424 { 425 426 if (l->l_class != SCHED_OTHER) 427 return; 428 429 KASSERT(!CURCPU_IDLE_P()); 430 l->l_estcpu = ESTCPULIM(l->l_estcpu + ESTCPU_ACCUM); 431 lwp_lock(l); 432 resetpriority(l); 433 lwp_unlock(l); 434 } 435 436 /* 437 * sched_proc_fork: 438 * 439 * Inherit the parent's scheduler history. 440 */ 441 void 442 sched_proc_fork(struct proc *parent, struct proc *child) 443 { 444 lwp_t *pl; 445 446 KASSERT(mutex_owned(parent->p_lock)); 447 448 pl = LIST_FIRST(&parent->p_lwps); 449 child->p_estcpu_inherited = pl->l_estcpu; 450 child->p_forktime = sched_pstats_ticks; 451 } 452 453 /* 454 * sched_proc_exit: 455 * 456 * Chargeback parents for the sins of their children. 457 */ 458 void 459 sched_proc_exit(struct proc *parent, struct proc *child) 460 { 461 fixpt_t loadfac = loadfactor(averunnable.ldavg[0]); 462 fixpt_t estcpu; 463 lwp_t *pl, *cl; 464 465 /* XXX Only if parent != init?? */ 466 467 mutex_enter(parent->p_lock); 468 pl = LIST_FIRST(&parent->p_lwps); 469 cl = LIST_FIRST(&child->p_lwps); 470 estcpu = decay_cpu_batch(loadfac, child->p_estcpu_inherited, 471 sched_pstats_ticks - child->p_forktime); 472 if (cl->l_estcpu > estcpu) { 473 lwp_lock(pl); 474 pl->l_estcpu = ESTCPULIM(pl->l_estcpu + cl->l_estcpu - estcpu); 475 lwp_unlock(pl); 476 } 477 mutex_exit(parent->p_lock); 478 } 479 480 void 481 sched_wakeup(struct lwp *l) 482 { 483 484 } 485 486 void 487 sched_slept(struct lwp *l) 488 { 489 490 } 491 492 void 493 sched_lwp_fork(struct lwp *l1, struct lwp *l2) 494 { 495 496 l2->l_estcpu = l1->l_estcpu; 497 } 498 499 void 500 sched_lwp_collect(struct lwp *t) 501 { 502 lwp_t *l; 503 504 /* Absorb estcpu value of collected LWP. */ 505 l = curlwp; 506 lwp_lock(l); 507 l->l_estcpu += t->l_estcpu; 508 lwp_unlock(l); 509 } 510 511 void 512 sched_oncpu(lwp_t *l) 513 { 514 515 } 516 517 void 518 sched_newts(lwp_t *l) 519 { 520 521 } 522 523 /* 524 * Sysctl nodes and initialization. 525 */ 526 527 static int 528 sysctl_sched_rtts(SYSCTLFN_ARGS) 529 { 530 struct sysctlnode node; 531 int rttsms = hztoms(rrticks); 532 533 node = *rnode; 534 node.sysctl_data = &rttsms; 535 return sysctl_lookup(SYSCTLFN_CALL(&node)); 536 } 537 538 SYSCTL_SETUP(sysctl_sched_4bsd_setup, "sysctl sched setup") 539 { 540 const struct sysctlnode *node = NULL; 541 542 sysctl_createv(clog, 0, NULL, &node, 543 CTLFLAG_PERMANENT, 544 CTLTYPE_NODE, "sched", 545 SYSCTL_DESCR("Scheduler options"), 546 NULL, 0, NULL, 0, 547 CTL_KERN, CTL_CREATE, CTL_EOL); 548 549 if (node == NULL) 550 return; 551 552 rrticks = hz / 10; 553 554 sysctl_createv(NULL, 0, &node, NULL, 555 CTLFLAG_PERMANENT, 556 CTLTYPE_STRING, "name", NULL, 557 NULL, 0, __UNCONST("4.4BSD"), 0, 558 CTL_CREATE, CTL_EOL); 559 sysctl_createv(NULL, 0, &node, NULL, 560 CTLFLAG_PERMANENT, 561 CTLTYPE_INT, "rtts", 562 SYSCTL_DESCR("Round-robin time quantum (in milliseconds)"), 563 sysctl_sched_rtts, 0, NULL, 0, 564 CTL_CREATE, CTL_EOL); 565 } 566
Indexes created Tue Sep 23 14:10:03 GMT 2025