sched_4bsd.c revision 1.22
1 /* $NetBSD: sched_4bsd.c,v 1.22 2008/05/19 12:48:54 rmind 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 1.22 2008/05/19 12:48:54 rmind 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/signalvar.h> 84 #include <sys/resourcevar.h> 85 #include <sys/sched.h> 86 #include <sys/sysctl.h> 87 #include <sys/kauth.h> 88 #include <sys/lockdebug.h> 89 #include <sys/kmem.h> 90 #include <sys/intr.h> 91 92 #include <uvm/uvm_extern.h> 93 94 static void updatepri(struct lwp *); 95 static void resetpriority(struct lwp *); 96 97 extern unsigned int sched_pstats_ticks; /* defined in kern_synch.c */ 98 99 /* Number of hardclock ticks per sched_tick() */ 100 static int rrticks; 101 102 /* 103 * Force switch among equal priority processes every 100ms. 104 * Called from hardclock every hz/10 == rrticks hardclock ticks. 105 * 106 * There's no need to lock anywhere in this routine, as it's 107 * CPU-local and runs at IPL_SCHED (called from clock interrupt). 108 */ 109 /* ARGSUSED */ 110 void 111 sched_tick(struct cpu_info *ci) 112 { 113 struct schedstate_percpu *spc = &ci->ci_schedstate; 114 115 spc->spc_ticks = rrticks; 116 117 if (CURCPU_IDLE_P()) { 118 cpu_need_resched(ci, 0); 119 return; 120 } 121 if (curlwp->l_class == SCHED_FIFO) { 122 return; 123 } 124 if (spc->spc_flags & SPCF_SEENRR) { 125 /* 126 * The process has already been through a roundrobin 127 * without switching and may be hogging the CPU. 128 * Indicate that the process should yield. 129 */ 130 spc->spc_flags |= SPCF_SHOULDYIELD; 131 cpu_need_resched(ci, 0); 132 } else 133 spc->spc_flags |= SPCF_SEENRR; 134 } 135 136 /* 137 * Why PRIO_MAX - 2? From setpriority(2): 138 * 139 * prio is a value in the range -20 to 20. The default priority is 140 * 0; lower priorities cause more favorable scheduling. A value of 141 * 19 or 20 will schedule a process only when nothing at priority <= 142 * 0 is runnable. 143 * 144 * This gives estcpu influence over 18 priority levels, and leaves nice 145 * with 40 levels. One way to think about it is that nice has 20 levels 146 * either side of estcpu's 18. 147 */ 148 #define ESTCPU_SHIFT 11 149 #define ESTCPU_MAX ((PRIO_MAX - 2) << ESTCPU_SHIFT) 150 #define ESTCPU_ACCUM (1 << (ESTCPU_SHIFT - 1)) 151 #define ESTCPULIM(e) min((e), ESTCPU_MAX) 152 153 /* 154 * Constants for digital decay and forget: 155 * 90% of (l_estcpu) usage in 5 * loadav time 156 * 95% of (l_pctcpu) usage in 60 seconds (load insensitive) 157 * Note that, as ps(1) mentions, this can let percentages 158 * total over 100% (I've seen 137.9% for 3 processes). 159 * 160 * Note that hardclock updates l_estcpu and l_cpticks independently. 161 * 162 * We wish to decay away 90% of l_estcpu in (5 * loadavg) seconds. 163 * That is, the system wants to compute a value of decay such 164 * that the following for loop: 165 * for (i = 0; i < (5 * loadavg); i++) 166 * l_estcpu *= decay; 167 * will compute 168 * l_estcpu *= 0.1; 169 * for all values of loadavg: 170 * 171 * Mathematically this loop can be expressed by saying: 172 * decay ** (5 * loadavg) ~= .1 173 * 174 * The system computes decay as: 175 * decay = (2 * loadavg) / (2 * loadavg + 1) 176 * 177 * We wish to prove that the system's computation of decay 178 * will always fulfill the equation: 179 * decay ** (5 * loadavg) ~= .1 180 * 181 * If we compute b as: 182 * b = 2 * loadavg 183 * then 184 * decay = b / (b + 1) 185 * 186 * We now need to prove two things: 187 * 1) Given factor ** (5 * loadavg) ~= .1, prove factor == b/(b+1) 188 * 2) Given b/(b+1) ** power ~= .1, prove power == (5 * loadavg) 189 * 190 * Facts: 191 * For x close to zero, exp(x) =~ 1 + x, since 192 * exp(x) = 0! + x**1/1! + x**2/2! + ... . 193 * therefore exp(-1/b) =~ 1 - (1/b) = (b-1)/b. 194 * For x close to zero, ln(1+x) =~ x, since 195 * ln(1+x) = x - x**2/2 + x**3/3 - ... -1 < x < 1 196 * therefore ln(b/(b+1)) = ln(1 - 1/(b+1)) =~ -1/(b+1). 197 * ln(.1) =~ -2.30 198 * 199 * Proof of (1): 200 * Solve (factor)**(power) =~ .1 given power (5*loadav): 201 * solving for factor, 202 * ln(factor) =~ (-2.30/5*loadav), or 203 * factor =~ exp(-1/((5/2.30)*loadav)) =~ exp(-1/(2*loadav)) = 204 * exp(-1/b) =~ (b-1)/b =~ b/(b+1). QED 205 * 206 * Proof of (2): 207 * Solve (factor)**(power) =~ .1 given factor == (b/(b+1)): 208 * solving for power, 209 * power*ln(b/(b+1)) =~ -2.30, or 210 * power =~ 2.3 * (b + 1) = 4.6*loadav + 2.3 =~ 5*loadav. QED 211 * 212 * Actual power values for the implemented algorithm are as follows: 213 * loadav: 1 2 3 4 214 * power: 5.68 10.32 14.94 19.55 215 */ 216 217 /* calculations for digital decay to forget 90% of usage in 5*loadav sec */ 218 #define loadfactor(loadav) (2 * (loadav)) 219 220 static fixpt_t 221 decay_cpu(fixpt_t loadfac, fixpt_t estcpu) 222 { 223 224 if (estcpu == 0) { 225 return 0; 226 } 227 228 #if !defined(_LP64) 229 /* avoid 64bit arithmetics. */ 230 #define FIXPT_MAX ((fixpt_t)((UINTMAX_C(1) << sizeof(fixpt_t) * CHAR_BIT) - 1)) 231 if (__predict_true(loadfac <= FIXPT_MAX / ESTCPU_MAX)) { 232 return estcpu * loadfac / (loadfac + FSCALE); 233 } 234 #endif /* !defined(_LP64) */ 235 236 return (uint64_t)estcpu * loadfac / (loadfac + FSCALE); 237 } 238 239 /* 240 * For all load averages >= 1 and max l_estcpu of (255 << ESTCPU_SHIFT), 241 * sleeping for at least seven times the loadfactor will decay l_estcpu to 242 * less than (1 << ESTCPU_SHIFT). 243 * 244 * note that our ESTCPU_MAX is actually much smaller than (255 << ESTCPU_SHIFT). 245 */ 246 static fixpt_t 247 decay_cpu_batch(fixpt_t loadfac, fixpt_t estcpu, unsigned int n) 248 { 249 250 if ((n << FSHIFT) >= 7 * loadfac) { 251 return 0; 252 } 253 254 while (estcpu != 0 && n > 1) { 255 estcpu = decay_cpu(loadfac, estcpu); 256 n--; 257 } 258 259 return estcpu; 260 } 261 262 /* 263 * sched_pstats_hook: 264 * 265 * Periodically called from sched_pstats(); used to recalculate priorities. 266 */ 267 void 268 sched_pstats_hook(struct lwp *l, int batch) 269 { 270 fixpt_t loadfac; 271 int sleeptm; 272 273 /* 274 * If the LWP has slept an entire second, stop recalculating 275 * its priority until it wakes up. 276 */ 277 if (l->l_stat == LSSLEEP || l->l_stat == LSSTOP || 278 l->l_stat == LSSUSPENDED) { 279 l->l_slptime++; 280 sleeptm = 1; 281 } else { 282 sleeptm = 0x7fffffff; 283 } 284 285 if (l->l_slptime <= sleeptm) { 286 loadfac = 2 * (averunnable.ldavg[0]); 287 l->l_estcpu = decay_cpu(loadfac, l->l_estcpu); 288 resetpriority(l); 289 } 290 } 291 292 /* 293 * Recalculate the priority of a process after it has slept for a while. 294 */ 295 static void 296 updatepri(struct lwp *l) 297 { 298 fixpt_t loadfac; 299 300 KASSERT(lwp_locked(l, NULL)); 301 KASSERT(l->l_slptime > 1); 302 303 loadfac = loadfactor(averunnable.ldavg[0]); 304 305 l->l_slptime--; /* the first time was done in sched_pstats */ 306 l->l_estcpu = decay_cpu_batch(loadfac, l->l_estcpu, l->l_slptime); 307 resetpriority(l); 308 } 309 310 void 311 sched_rqinit(void) 312 { 313 314 } 315 316 void 317 sched_setrunnable(struct lwp *l) 318 { 319 320 if (l->l_slptime > 1) 321 updatepri(l); 322 } 323 324 void 325 sched_nice(struct proc *p, int n) 326 { 327 struct lwp *l; 328 329 KASSERT(mutex_owned(p->p_lock)); 330 331 p->p_nice = n; 332 LIST_FOREACH(l, &p->p_lwps, l_sibling) { 333 lwp_lock(l); 334 resetpriority(l); 335 lwp_unlock(l); 336 } 337 } 338 339 /* 340 * Recompute the priority of an LWP. Arrange to reschedule if 341 * the resulting priority is better than that of the current LWP. 342 */ 343 static void 344 resetpriority(struct lwp *l) 345 { 346 pri_t pri; 347 struct proc *p = l->l_proc; 348 349 KASSERT(lwp_locked(l, NULL)); 350 351 if (l->l_class != SCHED_OTHER) 352 return; 353 354 /* See comments above ESTCPU_SHIFT definition. */ 355 pri = (PRI_KERNEL - 1) - (l->l_estcpu >> ESTCPU_SHIFT) - p->p_nice; 356 pri = imax(pri, 0); 357 if (pri != l->l_priority) 358 lwp_changepri(l, pri); 359 } 360 361 /* 362 * We adjust the priority of the current process. The priority of a process 363 * gets worse as it accumulates CPU time. The CPU usage estimator (l_estcpu) 364 * is increased here. The formula for computing priorities (in kern_synch.c) 365 * will compute a different value each time l_estcpu increases. This can 366 * cause a switch, but unless the priority crosses a PPQ boundary the actual 367 * queue will not change. The CPU usage estimator ramps up quite quickly 368 * when the process is running (linearly), and decays away exponentially, at 369 * a rate which is proportionally slower when the system is busy. The basic 370 * principle is that the system will 90% forget that the process used a lot 371 * of CPU time in 5 * loadav seconds. This causes the system to favor 372 * processes which haven't run much recently, and to round-robin among other 373 * processes. 374 */ 375 376 void 377 sched_schedclock(struct lwp *l) 378 { 379 380 if (l->l_class != SCHED_OTHER) 381 return; 382 383 KASSERT(!CURCPU_IDLE_P()); 384 l->l_estcpu = ESTCPULIM(l->l_estcpu + ESTCPU_ACCUM); 385 lwp_lock(l); 386 resetpriority(l); 387 lwp_unlock(l); 388 } 389 390 /* 391 * sched_proc_fork: 392 * 393 * Inherit the parent's scheduler history. 394 */ 395 void 396 sched_proc_fork(struct proc *parent, struct proc *child) 397 { 398 lwp_t *pl; 399 400 KASSERT(mutex_owned(parent->p_lock)); 401 402 pl = LIST_FIRST(&parent->p_lwps); 403 child->p_estcpu_inherited = pl->l_estcpu; 404 child->p_forktime = sched_pstats_ticks; 405 } 406 407 /* 408 * sched_proc_exit: 409 * 410 * Chargeback parents for the sins of their children. 411 */ 412 void 413 sched_proc_exit(struct proc *parent, struct proc *child) 414 { 415 fixpt_t loadfac = loadfactor(averunnable.ldavg[0]); 416 fixpt_t estcpu; 417 lwp_t *pl, *cl; 418 419 /* XXX Only if parent != init?? */ 420 421 mutex_enter(parent->p_lock); 422 pl = LIST_FIRST(&parent->p_lwps); 423 cl = LIST_FIRST(&child->p_lwps); 424 estcpu = decay_cpu_batch(loadfac, child->p_estcpu_inherited, 425 sched_pstats_ticks - child->p_forktime); 426 if (cl->l_estcpu > estcpu) { 427 lwp_lock(pl); 428 pl->l_estcpu = ESTCPULIM(pl->l_estcpu + cl->l_estcpu - estcpu); 429 lwp_unlock(pl); 430 } 431 mutex_exit(parent->p_lock); 432 } 433 434 void 435 sched_wakeup(struct lwp *l) 436 { 437 438 } 439 440 void 441 sched_slept(struct lwp *l) 442 { 443 444 } 445 446 void 447 sched_lwp_fork(struct lwp *l1, struct lwp *l2) 448 { 449 450 l2->l_estcpu = l1->l_estcpu; 451 } 452 453 void 454 sched_lwp_exit(struct lwp *l) 455 { 456 457 } 458 459 void 460 sched_lwp_collect(struct lwp *t) 461 { 462 lwp_t *l; 463 464 /* Absorb estcpu value of collected LWP. */ 465 l = curlwp; 466 lwp_lock(l); 467 l->l_estcpu += t->l_estcpu; 468 lwp_unlock(l); 469 } 470 471 void 472 sched_oncpu(lwp_t *l) 473 { 474 475 } 476 477 void 478 sched_newts(lwp_t *l) 479 { 480 481 } 482 483 /* 484 * Sysctl nodes and initialization. 485 */ 486 487 static int 488 sysctl_sched_rtts(SYSCTLFN_ARGS) 489 { 490 struct sysctlnode node; 491 int rttsms = hztoms(rrticks); 492 493 node = *rnode; 494 node.sysctl_data = &rttsms; 495 return sysctl_lookup(SYSCTLFN_CALL(&node)); 496 } 497 498 SYSCTL_SETUP(sysctl_sched_4bsd_setup, "sysctl sched setup") 499 { 500 const struct sysctlnode *node = NULL; 501 502 sysctl_createv(clog, 0, NULL, NULL, 503 CTLFLAG_PERMANENT, 504 CTLTYPE_NODE, "kern", NULL, 505 NULL, 0, NULL, 0, 506 CTL_KERN, CTL_EOL); 507 sysctl_createv(clog, 0, NULL, &node, 508 CTLFLAG_PERMANENT, 509 CTLTYPE_NODE, "sched", 510 SYSCTL_DESCR("Scheduler options"), 511 NULL, 0, NULL, 0, 512 CTL_KERN, CTL_CREATE, CTL_EOL); 513 514 if (node == NULL) 515 return; 516 517 rrticks = hz / 10; 518 519 sysctl_createv(NULL, 0, &node, NULL, 520 CTLFLAG_PERMANENT, 521 CTLTYPE_STRING, "name", NULL, 522 NULL, 0, __UNCONST("4.4BSD"), 0, 523 CTL_CREATE, CTL_EOL); 524 sysctl_createv(NULL, 0, &node, NULL, 525 CTLFLAG_PERMANENT, 526 CTLTYPE_INT, "rtts", 527 SYSCTL_DESCR("Round-robin time quantum (in miliseconds)"), 528 sysctl_sched_rtts, 0, NULL, 0, 529 CTL_CREATE, CTL_EOL); 530 } 531
Indexes created Wed Sep 24 05:09:52 GMT 2025