1 /* $NetBSD: sched_4bsd.c,v 1.47 2025/01/17 04:11:33 mrg Exp $ */ 2 3 /* 4 * Copyright (c) 1999, 2000, 2004, 2006, 2007, 2008, 2019, 2020 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 1.47 2025/01/17 04:11:33 mrg 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 /* Number of hardclock ticks per sched_tick() */ 94 u_int sched_rrticks __read_mostly; 95 96 /* 97 * Force switch among equal priority processes every 100ms. 98 * Called from hardclock every hz/10 == sched_rrticks hardclock ticks. 99 */ 100 /* ARGSUSED */ 101 void 102 sched_tick(struct cpu_info *ci) 103 { 104 struct schedstate_percpu *spc = &ci->ci_schedstate; 105 pri_t pri = PRI_NONE; 106 lwp_t *l; 107 108 spc->spc_ticks = sched_rrticks; 109 110 if (CURCPU_IDLE_P()) { 111 spc_lock(ci); 112 sched_resched_cpu(ci, MAXPRI_KTHREAD, true); 113 /* spc now unlocked */ 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 pri = MAXPRI_KERNEL_RT; 132 break; 133 default: 134 if (spc->spc_flags & SPCF_SHOULDYIELD) { 135 /* 136 * Process is stuck in kernel somewhere, probably 137 * due to buggy or inefficient code. Force a 138 * kernel preemption. 139 */ 140 pri = MAXPRI_KERNEL_RT; 141 } else if (spc->spc_flags & SPCF_SEENRR) { 142 /* 143 * The process has already been through a roundrobin 144 * without switching and may be hogging the CPU. 145 * Indicate that the process should yield. 146 */ 147 pri = MAXPRI_KTHREAD; 148 spc->spc_flags |= SPCF_SHOULDYIELD; 149 } else if (!cpu_is_1stclass(ci)) { 150 /* 151 * For SMT or asymmetric systems push a little 152 * harder: if this is not a 1st class CPU, try to 153 * find a better one to run this LWP. 154 */ 155 pri = MAXPRI_KTHREAD; 156 spc->spc_flags |= SPCF_SHOULDYIELD; 157 } else { 158 spc->spc_flags |= SPCF_SEENRR; 159 } 160 break; 161 } 162 163 if (pri != PRI_NONE) { 164 spc_lock(ci); 165 sched_resched_cpu(ci, pri, true); 166 /* spc now unlocked */ 167 } 168 } 169 170 /* 171 * Why PRIO_MAX - 2? From setpriority(2): 172 * 173 * prio is a value in the range -20 to 20. The default priority is 174 * 0; lower priorities cause more favorable scheduling. A value of 175 * 19 or 20 will schedule a process only when nothing at priority <= 176 * 0 is runnable. 177 * 178 * This gives estcpu influence over 18 priority levels, and leaves nice 179 * with 40 levels. One way to think about it is that nice has 20 levels 180 * either side of estcpu's 18. 181 */ 182 #define ESTCPU_SHIFT 11 183 #define ESTCPU_MAX ((PRIO_MAX - 2) << ESTCPU_SHIFT) 184 #define ESTCPU_ACCUM (1 << (ESTCPU_SHIFT - 1)) 185 #define ESTCPULIM(e) uimin((e), ESTCPU_MAX) 186 187 /* 188 * The main parameter used by this algorithm is 'l_estcpu'. It is an estimate 189 * of the recent CPU utilization of the thread. 190 * 191 * l_estcpu is: 192 * - increased each time the hardclock ticks and the thread is found to 193 * be executing, in sched_schedclock() called from hardclock() 194 * - decreased (filtered) on each sched tick, in sched_pstats_hook() 195 * If the lwp is sleeping for more than a second, we don't touch l_estcpu: it 196 * will be updated in sched_setrunnable() when the lwp wakes up, in burst mode 197 * (ie, we decrease it n times). 198 * 199 * Note that hardclock updates l_estcpu and l_cpticks independently. 200 * 201 * ----------------------------------------------------------------------------- 202 * 203 * Here we describe how l_estcpu is decreased. 204 * 205 * Constants for digital decay (filter): 206 * 90% of l_estcpu usage in (5 * loadavg) seconds 207 * 208 * We wish to decay away 90% of l_estcpu in (5 * loadavg) seconds. That is, we 209 * want to compute a value of decay such that the following loop: 210 * for (i = 0; i < (5 * loadavg); i++) 211 * l_estcpu *= decay; 212 * will result in 213 * l_estcpu *= 0.1; 214 * for all values of loadavg. 215 * 216 * Mathematically this loop can be expressed by saying: 217 * decay ** (5 * loadavg) ~= .1 218 * 219 * And finally, the corresponding value of decay we're using is: 220 * decay = (2 * loadavg) / (2 * loadavg + 1) 221 * 222 * ----------------------------------------------------------------------------- 223 * 224 * Now, let's prove that the value of decay stated above will always fulfill 225 * the equation: 226 * decay ** (5 * loadavg) ~= .1 227 * 228 * If we compute b as: 229 * b = 2 * loadavg 230 * then 231 * decay = b / (b + 1) 232 * 233 * We now need to prove two things: 234 * 1) Given [factor ** (5 * loadavg) =~ .1], prove [factor == b/(b+1)]. 235 * 2) Given [b/(b+1) ** power =~ .1], prove [power == (5 * loadavg)]. 236 * 237 * Facts: 238 * * For x real: exp(x) = 0! + x**1/1! + x**2/2! + ... 239 * Therefore, for x close to zero, exp(x) =~ 1 + x. 240 * In turn, for b large enough, exp(-1/b) =~ 1 - (1/b) = (b-1)/b. 241 * 242 * * For b large enough, (b-1)/b =~ b/(b+1). 243 * 244 * * For x belonging to [-1;1[, ln(1-x) = - x - x**2/2 - x**3/3 - ... 245 * Therefore ln(b/(b+1)) = ln(1 - 1/(b+1)) =~ -1/(b+1). 246 * 247 * * ln(0.1) =~ -2.30 248 * 249 * Proof of (1): 250 * factor ** (5 * loadavg) =~ 0.1 251 * => ln(factor) =~ -2.30 / (5 * loadavg) 252 * => factor =~ exp(-1 / ((5 / 2.30) * loadavg)) 253 * =~ exp(-1 / (2 * loadavg)) 254 * =~ exp(-1 / b) 255 * =~ (b - 1) / b 256 * =~ b / (b + 1) 257 * =~ (2 * loadavg) / ((2 * loadavg) + 1) 258 * 259 * Proof of (2): 260 * (b / (b + 1)) ** power =~ .1 261 * => power * ln(b / (b + 1)) =~ -2.30 262 * => power * (-1 / (b + 1)) =~ -2.30 263 * => power =~ 2.30 * (b + 1) 264 * => power =~ 4.60 * loadavg + 2.30 265 * => power =~ 5 * loadavg 266 * 267 * Conclusion: decay = (2 * loadavg) / (2 * loadavg + 1) 268 */ 269 270 /* See calculations above */ 271 #define loadfactor(loadavg) (2 * (loadavg)) 272 273 static fixpt_t 274 decay_cpu(fixpt_t loadfac, fixpt_t estcpu) 275 { 276 277 if (estcpu == 0) { 278 return 0; 279 } 280 281 #if !defined(_LP64) 282 /* avoid 64bit arithmetics. */ 283 #define FIXPT_MAX ((fixpt_t)((UINTMAX_C(1) << sizeof(fixpt_t) * CHAR_BIT) - 1)) 284 if (__predict_true(loadfac <= FIXPT_MAX / ESTCPU_MAX)) { 285 return estcpu * loadfac / (loadfac + FSCALE); 286 } 287 #endif 288 289 return (uint64_t)estcpu * loadfac / (loadfac + FSCALE); 290 } 291 292 static fixpt_t 293 decay_cpu_batch(fixpt_t loadfac, fixpt_t estcpu, unsigned int n) 294 { 295 296 /* 297 * For all load averages >= 1 and max l_estcpu of (255 << ESTCPU_SHIFT), 298 * if we slept for at least seven times the loadfactor, we will decay 299 * l_estcpu to less than (1 << ESTCPU_SHIFT), and therefore we can 300 * return zero directly. 301 * 302 * Note that our ESTCPU_MAX is actually much smaller than 303 * (255 << ESTCPU_SHIFT). 304 */ 305 if ((n << FSHIFT) >= 7 * loadfac) { 306 return 0; 307 } 308 309 while (estcpu != 0 && n > 1) { 310 estcpu = decay_cpu(loadfac, estcpu); 311 n--; 312 } 313 314 return estcpu; 315 } 316 317 /* 318 * sched_pstats_hook: 319 * 320 * Periodically called from sched_pstats(); used to recalculate priorities. 321 */ 322 void 323 sched_pstats_hook(struct lwp *l, int batch) 324 { 325 fixpt_t loadfac; 326 327 /* 328 * If the LWP has slept an entire second, stop recalculating 329 * its priority until it wakes up. 330 */ 331 KASSERT(lwp_locked(l, NULL)); 332 if (l->l_stat == LSSLEEP || l->l_stat == LSSTOP || 333 l->l_stat == LSSUSPENDED) { 334 if (l->l_slptime > 1) { 335 return; 336 } 337 } 338 339 loadfac = loadfactor(averunnable.ldavg[0]); 340 l->l_estcpu = decay_cpu(loadfac, l->l_estcpu); 341 resetpriority(l); 342 } 343 344 /* 345 * Recalculate the priority of an LWP after it has slept for a while. 346 */ 347 static void 348 updatepri(struct lwp *l) 349 { 350 fixpt_t loadfac; 351 352 KASSERT(lwp_locked(l, NULL)); 353 KASSERT(l->l_slptime > 1); 354 355 loadfac = loadfactor(averunnable.ldavg[0]); 356 357 l->l_slptime--; /* the first time was done in sched_pstats */ 358 l->l_estcpu = decay_cpu_batch(loadfac, l->l_estcpu, l->l_slptime); 359 resetpriority(l); 360 } 361 362 void 363 sched_rqinit(void) 364 { 365 366 } 367 368 void 369 sched_setrunnable(struct lwp *l) 370 { 371 372 if (l->l_slptime > 1) 373 updatepri(l); 374 } 375 376 void 377 sched_nice(struct proc *p, int n) 378 { 379 struct lwp *l; 380 381 KASSERT(mutex_owned(p->p_lock)); 382 383 p->p_nice = n; 384 LIST_FOREACH(l, &p->p_lwps, l_sibling) { 385 lwp_lock(l); 386 resetpriority(l); 387 lwp_unlock(l); 388 } 389 } 390 391 /* 392 * Recompute the priority of an LWP. Arrange to reschedule if 393 * the resulting priority is better than that of the current LWP. 394 */ 395 static void 396 resetpriority(struct lwp *l) 397 { 398 pri_t pri; 399 struct proc *p = l->l_proc; 400 401 KASSERT(lwp_locked(l, NULL)); 402 403 if (l->l_class != SCHED_OTHER) 404 return; 405 406 /* See comments above ESTCPU_SHIFT definition. */ 407 pri = (PRI_KERNEL - 1) - (l->l_estcpu >> ESTCPU_SHIFT) - p->p_nice; 408 pri = imax(pri, 0); 409 if (pri != l->l_priority) 410 lwp_changepri(l, pri); 411 } 412 413 /* 414 * We adjust the priority of the current LWP. The priority of a LWP 415 * gets worse as it accumulates CPU time. The CPU usage estimator (l_estcpu) 416 * is increased here. The formula for computing priorities will compute a 417 * different value each time l_estcpu increases. This can cause a switch, 418 * but unless the priority crosses a PPQ boundary the actual queue will not 419 * change. The CPU usage estimator ramps up quite quickly when the process 420 * is running (linearly), and decays away exponentially, at a rate which is 421 * proportionally slower when the system is busy. The basic principle is 422 * that the system will 90% forget that the process used a lot of CPU time 423 * in (5 * loadavg) seconds. This causes the system to favor processes which 424 * haven't run much recently, and to round-robin among other processes. 425 */ 426 void 427 sched_schedclock(struct lwp *l) 428 { 429 430 if (l->l_class != SCHED_OTHER) 431 return; 432 433 KASSERT(!CURCPU_IDLE_P()); 434 l->l_estcpu = ESTCPULIM(l->l_estcpu + ESTCPU_ACCUM); 435 lwp_lock(l); 436 resetpriority(l); 437 lwp_unlock(l); 438 } 439 440 /* 441 * sched_proc_fork: 442 * 443 * Inherit the parent's scheduler history. 444 */ 445 void 446 sched_proc_fork(struct proc *parent, struct proc *child) 447 { 448 lwp_t *pl; 449 450 KASSERT(mutex_owned(parent->p_lock)); 451 452 pl = LIST_FIRST(&parent->p_lwps); 453 child->p_estcpu_inherited = pl->l_estcpu; 454 child->p_forktime = sched_pstats_ticks; 455 } 456 457 /* 458 * sched_proc_exit: 459 * 460 * Chargeback parents for the sins of their children. 461 */ 462 void 463 sched_proc_exit(struct proc *parent, struct proc *child) 464 { 465 fixpt_t loadfac = loadfactor(averunnable.ldavg[0]); 466 fixpt_t estcpu; 467 lwp_t *pl, *cl; 468 469 /* XXX Only if parent != init?? */ 470 471 mutex_enter(parent->p_lock); 472 pl = LIST_FIRST(&parent->p_lwps); 473 cl = LIST_FIRST(&child->p_lwps); 474 estcpu = decay_cpu_batch(loadfac, child->p_estcpu_inherited, 475 sched_pstats_ticks - child->p_forktime); 476 if (cl->l_estcpu > estcpu) { 477 lwp_lock(pl); 478 pl->l_estcpu = ESTCPULIM(pl->l_estcpu + cl->l_estcpu - estcpu); 479 lwp_unlock(pl); 480 } 481 mutex_exit(parent->p_lock); 482 } 483 484 void 485 sched_wakeup(struct lwp *l) 486 { 487 488 } 489 490 void 491 sched_slept(struct lwp *l) 492 { 493 494 } 495 496 void 497 sched_lwp_fork(struct lwp *l1, struct lwp *l2) 498 { 499 500 l2->l_estcpu = l1->l_estcpu; 501 } 502 503 void 504 sched_lwp_collect(struct lwp *t) 505 { 506 lwp_t *l; 507 508 /* Absorb estcpu value of collected LWP. */ 509 l = curlwp; 510 lwp_lock(l); 511 l->l_estcpu += t->l_estcpu; 512 lwp_unlock(l); 513 } 514 515 void 516 sched_oncpu(lwp_t *l) 517 { 518 519 } 520 521 void 522 sched_newts(lwp_t *l) 523 { 524 525 } 526 527 /* 528 * Sysctl nodes and initialization. 529 */ 530 531 static int 532 sysctl_sched_rtts(SYSCTLFN_ARGS) 533 { 534 struct sysctlnode node; 535 int rttsms = hztoms(sched_rrticks); 536 537 node = *rnode; 538 node.sysctl_data = &rttsms; 539 return sysctl_lookup(SYSCTLFN_CALL(&node)); 540 } 541 542 SYSCTL_SETUP(sysctl_sched_4bsd_setup, "sysctl sched setup") 543 { 544 const struct sysctlnode *node = NULL; 545 546 sysctl_createv(clog, 0, NULL, &node, 547 CTLFLAG_PERMANENT, 548 CTLTYPE_NODE, "sched", 549 SYSCTL_DESCR("Scheduler options"), 550 NULL, 0, NULL, 0, 551 CTL_KERN, CTL_CREATE, CTL_EOL); 552 553 if (node == NULL) 554 return; 555 556 sched_rrticks = hz / 10; 557 558 sysctl_createv(NULL, 0, &node, NULL, 559 CTLFLAG_PERMANENT, 560 CTLTYPE_STRING, "name", NULL, 561 NULL, 0, __UNCONST("4.4BSD"), 0, 562 CTL_CREATE, CTL_EOL); 563 sysctl_createv(NULL, 0, &node, NULL, 564 CTLFLAG_PERMANENT, 565 CTLTYPE_INT, "rtts", 566 SYSCTL_DESCR("Round-robin time quantum (in milliseconds)"), 567 sysctl_sched_rtts, 0, NULL, 0, 568 CTL_CREATE, CTL_EOL); 569 } 570
Indexes created Mon Sep 22 05:09:51 GMT 2025