1 /* $NetBSD: sched_4bsd.c,v 1.4 2007/08/04 11:03:02 ad Exp $ */ 2 3 /*- 4 * Copyright (c) 1999, 2000, 2004, 2006, 2007 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 * 3. All advertising materials mentioning features or use of this software 21 * must display the following acknowledgement: 22 * This product includes software developed by the NetBSD 23 * Foundation, Inc. and its contributors. 24 * 4. Neither the name of The NetBSD Foundation nor the names of its 25 * contributors may be used to endorse or promote products derived 26 * from this software without specific prior written permission. 27 * 28 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 29 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 30 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 31 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 32 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 33 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 34 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 35 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 36 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 37 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 38 * POSSIBILITY OF SUCH DAMAGE. 39 */ 40 41 /*- 42 * Copyright (c) 1982, 1986, 1990, 1991, 1993 43 * The Regents of the University of California. All rights reserved. 44 * (c) UNIX System Laboratories, Inc. 45 * All or some portions of this file are derived from material licensed 46 * to the University of California by American Telephone and Telegraph 47 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 48 * the permission of UNIX System Laboratories, Inc. 49 * 50 * Redistribution and use in source and binary forms, with or without 51 * modification, are permitted provided that the following conditions 52 * are met: 53 * 1. Redistributions of source code must retain the above copyright 54 * notice, this list of conditions and the following disclaimer. 55 * 2. Redistributions in binary form must reproduce the above copyright 56 * notice, this list of conditions and the following disclaimer in the 57 * documentation and/or other materials provided with the distribution. 58 * 3. Neither the name of the University nor the names of its contributors 59 * may be used to endorse or promote products derived from this software 60 * without specific prior written permission. 61 * 62 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 63 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 64 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 65 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 66 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 67 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 68 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 69 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 70 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 71 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 72 * SUCH DAMAGE. 73 * 74 * @(#)kern_synch.c 8.9 (Berkeley) 5/19/95 75 */ 76 77 #include <sys/cdefs.h> 78 __KERNEL_RCSID(0, "$NetBSD: sched_4bsd.c,v 1.4 2007/08/04 11:03:02 ad Exp $"); 79 80 #include "opt_ddb.h" 81 #include "opt_lockdebug.h" 82 #include "opt_perfctrs.h" 83 84 #define __MUTEX_PRIVATE 85 86 #include <sys/param.h> 87 #include <sys/systm.h> 88 #include <sys/callout.h> 89 #include <sys/cpu.h> 90 #include <sys/proc.h> 91 #include <sys/kernel.h> 92 #include <sys/signalvar.h> 93 #include <sys/resourcevar.h> 94 #include <sys/sched.h> 95 #include <sys/sysctl.h> 96 #include <sys/kauth.h> 97 #include <sys/lockdebug.h> 98 #include <sys/kmem.h> 99 100 #include <uvm/uvm_extern.h> 101 102 /* 103 * Run queues. 104 * 105 * We have 32 run queues in descending priority of 0..31. We maintain 106 * a bitmask of non-empty queues in order speed up finding the first 107 * runnable process. The bitmask is maintained only by machine-dependent 108 * code, allowing the most efficient instructions to be used to find the 109 * first non-empty queue. 110 */ 111 112 #define RUNQUE_NQS 32 /* number of runqueues */ 113 #define PPQ (128 / RUNQUE_NQS) /* priorities per queue */ 114 115 typedef struct subqueue { 116 TAILQ_HEAD(, lwp) sq_queue; 117 } subqueue_t; 118 typedef struct runqueue { 119 subqueue_t rq_subqueues[RUNQUE_NQS]; /* run queues */ 120 uint32_t rq_bitmap; /* bitmap of non-empty queues */ 121 } runqueue_t; 122 static runqueue_t global_queue; 123 124 static void updatepri(struct lwp *); 125 static void resetpriority(struct lwp *); 126 static void resetprocpriority(struct proc *); 127 128 extern unsigned int sched_pstats_ticks; /* defined in kern_synch.c */ 129 130 /* The global scheduler state */ 131 kmutex_t sched_mutex; 132 133 /* Number of hardclock ticks per sched_tick() */ 134 int rrticks; 135 136 /* 137 * Force switch among equal priority processes every 100ms. 138 * Called from hardclock every hz/10 == rrticks hardclock ticks. 139 */ 140 /* ARGSUSED */ 141 void 142 sched_tick(struct cpu_info *ci) 143 { 144 struct schedstate_percpu *spc = &ci->ci_schedstate; 145 146 spc->spc_ticks = rrticks; 147 148 spc_lock(ci); 149 if (!CURCPU_IDLE_P()) { 150 if (spc->spc_flags & SPCF_SEENRR) { 151 /* 152 * The process has already been through a roundrobin 153 * without switching and may be hogging the CPU. 154 * Indicate that the process should yield. 155 */ 156 spc->spc_flags |= SPCF_SHOULDYIELD; 157 } else 158 spc->spc_flags |= SPCF_SEENRR; 159 } 160 cpu_need_resched(curcpu(), 0); 161 spc_unlock(ci); 162 } 163 164 #define NICE_WEIGHT 2 /* priorities per nice level */ 165 166 #define ESTCPU_SHIFT 11 167 #define ESTCPU_MAX ((NICE_WEIGHT * PRIO_MAX - PPQ) << ESTCPU_SHIFT) 168 #define ESTCPULIM(e) min((e), ESTCPU_MAX) 169 170 /* 171 * Constants for digital decay and forget: 172 * 90% of (p_estcpu) usage in 5 * loadav time 173 * 95% of (p_pctcpu) usage in 60 seconds (load insensitive) 174 * Note that, as ps(1) mentions, this can let percentages 175 * total over 100% (I've seen 137.9% for 3 processes). 176 * 177 * Note that hardclock updates p_estcpu and p_cpticks independently. 178 * 179 * We wish to decay away 90% of p_estcpu in (5 * loadavg) seconds. 180 * That is, the system wants to compute a value of decay such 181 * that the following for loop: 182 * for (i = 0; i < (5 * loadavg); i++) 183 * p_estcpu *= decay; 184 * will compute 185 * p_estcpu *= 0.1; 186 * for all values of loadavg: 187 * 188 * Mathematically this loop can be expressed by saying: 189 * decay ** (5 * loadavg) ~= .1 190 * 191 * The system computes decay as: 192 * decay = (2 * loadavg) / (2 * loadavg + 1) 193 * 194 * We wish to prove that the system's computation of decay 195 * will always fulfill the equation: 196 * decay ** (5 * loadavg) ~= .1 197 * 198 * If we compute b as: 199 * b = 2 * loadavg 200 * then 201 * decay = b / (b + 1) 202 * 203 * We now need to prove two things: 204 * 1) Given factor ** (5 * loadavg) ~= .1, prove factor == b/(b+1) 205 * 2) Given b/(b+1) ** power ~= .1, prove power == (5 * loadavg) 206 * 207 * Facts: 208 * For x close to zero, exp(x) =~ 1 + x, since 209 * exp(x) = 0! + x**1/1! + x**2/2! + ... . 210 * therefore exp(-1/b) =~ 1 - (1/b) = (b-1)/b. 211 * For x close to zero, ln(1+x) =~ x, since 212 * ln(1+x) = x - x**2/2 + x**3/3 - ... -1 < x < 1 213 * therefore ln(b/(b+1)) = ln(1 - 1/(b+1)) =~ -1/(b+1). 214 * ln(.1) =~ -2.30 215 * 216 * Proof of (1): 217 * Solve (factor)**(power) =~ .1 given power (5*loadav): 218 * solving for factor, 219 * ln(factor) =~ (-2.30/5*loadav), or 220 * factor =~ exp(-1/((5/2.30)*loadav)) =~ exp(-1/(2*loadav)) = 221 * exp(-1/b) =~ (b-1)/b =~ b/(b+1). QED 222 * 223 * Proof of (2): 224 * Solve (factor)**(power) =~ .1 given factor == (b/(b+1)): 225 * solving for power, 226 * power*ln(b/(b+1)) =~ -2.30, or 227 * power =~ 2.3 * (b + 1) = 4.6*loadav + 2.3 =~ 5*loadav. QED 228 * 229 * Actual power values for the implemented algorithm are as follows: 230 * loadav: 1 2 3 4 231 * power: 5.68 10.32 14.94 19.55 232 */ 233 234 /* calculations for digital decay to forget 90% of usage in 5*loadav sec */ 235 #define loadfactor(loadav) (2 * (loadav)) 236 237 static fixpt_t 238 decay_cpu(fixpt_t loadfac, fixpt_t estcpu) 239 { 240 241 if (estcpu == 0) { 242 return 0; 243 } 244 245 #if !defined(_LP64) 246 /* avoid 64bit arithmetics. */ 247 #define FIXPT_MAX ((fixpt_t)((UINTMAX_C(1) << sizeof(fixpt_t) * CHAR_BIT) - 1)) 248 if (__predict_true(loadfac <= FIXPT_MAX / ESTCPU_MAX)) { 249 return estcpu * loadfac / (loadfac + FSCALE); 250 } 251 #endif /* !defined(_LP64) */ 252 253 return (uint64_t)estcpu * loadfac / (loadfac + FSCALE); 254 } 255 256 /* 257 * For all load averages >= 1 and max p_estcpu of (255 << ESTCPU_SHIFT), 258 * sleeping for at least seven times the loadfactor will decay p_estcpu to 259 * less than (1 << ESTCPU_SHIFT). 260 * 261 * note that our ESTCPU_MAX is actually much smaller than (255 << ESTCPU_SHIFT). 262 */ 263 static fixpt_t 264 decay_cpu_batch(fixpt_t loadfac, fixpt_t estcpu, unsigned int n) 265 { 266 267 if ((n << FSHIFT) >= 7 * loadfac) { 268 return 0; 269 } 270 271 while (estcpu != 0 && n > 1) { 272 estcpu = decay_cpu(loadfac, estcpu); 273 n--; 274 } 275 276 return estcpu; 277 } 278 279 /* 280 * sched_pstats_hook: 281 * 282 * Periodically called from sched_pstats(); used to recalculate priorities. 283 */ 284 void 285 sched_pstats_hook(struct proc *p, int minslp) 286 { 287 struct lwp *l; 288 fixpt_t loadfac = loadfactor(averunnable.ldavg[0]); 289 290 /* 291 * If the process has slept the entire second, 292 * stop recalculating its priority until it wakes up. 293 */ 294 if (minslp <= 1) { 295 p->p_estcpu = decay_cpu(loadfac, p->p_estcpu); 296 297 LIST_FOREACH(l, &p->p_lwps, l_sibling) { 298 if ((l->l_flag & LW_IDLE) != 0) 299 continue; 300 lwp_lock(l); 301 if (l->l_slptime <= 1 && l->l_priority >= PUSER) 302 resetpriority(l); 303 lwp_unlock(l); 304 } 305 } 306 } 307 308 /* 309 * Recalculate the priority of a process after it has slept for a while. 310 */ 311 static void 312 updatepri(struct lwp *l) 313 { 314 struct proc *p = l->l_proc; 315 fixpt_t loadfac; 316 317 KASSERT(lwp_locked(l, NULL)); 318 KASSERT(l->l_slptime > 1); 319 320 loadfac = loadfactor(averunnable.ldavg[0]); 321 322 l->l_slptime--; /* the first time was done in sched_pstats */ 323 /* XXX NJWLWP */ 324 /* XXXSMP occasionally unlocked, should be per-LWP */ 325 p->p_estcpu = decay_cpu_batch(loadfac, p->p_estcpu, l->l_slptime); 326 resetpriority(l); 327 } 328 329 /* 330 * On some architectures, it's faster to use a MSB ordering for the priorites 331 * than the traditional LSB ordering. 332 */ 333 #define RQMASK(n) (0x00000001 << (n)) 334 335 /* 336 * The primitives that manipulate the run queues. whichqs tells which 337 * of the 32 queues qs have processes in them. sched_enqueue() puts processes 338 * into queues, sched_dequeue removes them from queues. The running process is 339 * on no queue, other processes are on a queue related to p->p_priority, 340 * divided by 4 actually to shrink the 0-127 range of priorities into the 32 341 * available queues. 342 */ 343 #ifdef RQDEBUG 344 static void 345 runqueue_check(const runqueue_t *rq, int whichq, struct lwp *l) 346 { 347 const subqueue_t * const sq = &rq->rq_subqueues[whichq]; 348 const uint32_t bitmap = rq->rq_bitmap; 349 struct lwp *l2; 350 int found = 0; 351 int die = 0; 352 int empty = 1; 353 354 TAILQ_FOREACH(l2, &sq->sq_queue, l_runq) { 355 if (l2->l_stat != LSRUN) { 356 printf("runqueue_check[%d]: lwp %p state (%d) " 357 " != LSRUN\n", whichq, l2, l2->l_stat); 358 } 359 if (l2 == l) 360 found = 1; 361 empty = 0; 362 } 363 if (empty && (bitmap & RQMASK(whichq)) != 0) { 364 printf("runqueue_check[%d]: bit set for empty run-queue %p\n", 365 whichq, rq); 366 die = 1; 367 } else if (!empty && (bitmap & RQMASK(whichq)) == 0) { 368 printf("runqueue_check[%d]: bit clear for non-empty " 369 "run-queue %p\n", whichq, rq); 370 die = 1; 371 } 372 if (l != NULL && (bitmap & RQMASK(whichq)) == 0) { 373 printf("runqueue_check[%d]: bit clear for active lwp %p\n", 374 whichq, l); 375 die = 1; 376 } 377 if (l != NULL && empty) { 378 printf("runqueue_check[%d]: empty run-queue %p with " 379 "active lwp %p\n", whichq, rq, l); 380 die = 1; 381 } 382 if (l != NULL && !found) { 383 printf("runqueue_check[%d]: lwp %p not in runqueue %p!", 384 whichq, l, rq); 385 die = 1; 386 } 387 if (die) 388 panic("runqueue_check: inconsistency found"); 389 } 390 #else /* RQDEBUG */ 391 #define runqueue_check(a, b, c) /* nothing */ 392 #endif /* RQDEBUG */ 393 394 static void 395 runqueue_init(runqueue_t *rq) 396 { 397 int i; 398 399 for (i = 0; i < RUNQUE_NQS; i++) 400 TAILQ_INIT(&rq->rq_subqueues[i].sq_queue); 401 } 402 403 static void 404 runqueue_enqueue(runqueue_t *rq, struct lwp *l) 405 { 406 subqueue_t *sq; 407 const int whichq = lwp_eprio(l) / PPQ; 408 409 KASSERT(lwp_locked(l, l->l_cpu->ci_schedstate.spc_mutex)); 410 411 runqueue_check(rq, whichq, NULL); 412 rq->rq_bitmap |= RQMASK(whichq); 413 sq = &rq->rq_subqueues[whichq]; 414 TAILQ_INSERT_TAIL(&sq->sq_queue, l, l_runq); 415 runqueue_check(rq, whichq, l); 416 } 417 418 static void 419 runqueue_dequeue(runqueue_t *rq, struct lwp *l) 420 { 421 subqueue_t *sq; 422 const int whichq = lwp_eprio(l) / PPQ; 423 424 KASSERT(lwp_locked(l, l->l_cpu->ci_schedstate.spc_mutex)); 425 426 runqueue_check(rq, whichq, l); 427 KASSERT((rq->rq_bitmap & RQMASK(whichq)) != 0); 428 sq = &rq->rq_subqueues[whichq]; 429 TAILQ_REMOVE(&sq->sq_queue, l, l_runq); 430 if (TAILQ_EMPTY(&sq->sq_queue)) 431 rq->rq_bitmap &= ~RQMASK(whichq); 432 runqueue_check(rq, whichq, NULL); 433 } 434 435 static struct lwp * 436 runqueue_nextlwp(runqueue_t *rq) 437 { 438 const uint32_t bitmap = rq->rq_bitmap; 439 int whichq; 440 441 if (bitmap == 0) { 442 return NULL; 443 } 444 whichq = ffs(bitmap) - 1; 445 return TAILQ_FIRST(&rq->rq_subqueues[whichq].sq_queue); 446 } 447 448 #if defined(DDB) 449 static void 450 runqueue_print(const runqueue_t *rq, void (*pr)(const char *, ...)) 451 { 452 const uint32_t bitmap = rq->rq_bitmap; 453 struct lwp *l; 454 int i, first; 455 456 for (i = 0; i < RUNQUE_NQS; i++) { 457 const subqueue_t *sq; 458 first = 1; 459 sq = &rq->rq_subqueues[i]; 460 TAILQ_FOREACH(l, &sq->sq_queue, l_runq) { 461 if (first) { 462 (*pr)("%c%d", 463 (bitmap & RQMASK(i)) ? ' ' : '!', i); 464 first = 0; 465 } 466 (*pr)("\t%d.%d (%s) pri=%d usrpri=%d\n", 467 l->l_proc->p_pid, 468 l->l_lid, l->l_proc->p_comm, 469 (int)l->l_priority, (int)l->l_usrpri); 470 } 471 } 472 } 473 #endif /* defined(DDB) */ 474 #undef RQMASK 475 476 /* 477 * Initialize the (doubly-linked) run queues 478 * to be empty. 479 */ 480 void 481 sched_rqinit() 482 { 483 484 runqueue_init(&global_queue); 485 mutex_init(&sched_mutex, MUTEX_SPIN, IPL_SCHED); 486 /* Initialize the lock pointer for lwp0 */ 487 lwp0.l_mutex = &curcpu()->ci_schedstate.spc_lwplock; 488 } 489 490 void 491 sched_cpuattach(struct cpu_info *ci) 492 { 493 runqueue_t *rq; 494 495 ci->ci_schedstate.spc_mutex = &sched_mutex; 496 rq = kmem_zalloc(sizeof(*rq), KM_NOSLEEP); 497 runqueue_init(rq); 498 ci->ci_schedstate.spc_sched_info = rq; 499 } 500 501 void 502 sched_setup() 503 { 504 505 rrticks = hz / 10; 506 } 507 508 void 509 sched_setrunnable(struct lwp *l) 510 { 511 512 if (l->l_slptime > 1) 513 updatepri(l); 514 } 515 516 bool 517 sched_curcpu_runnable_p(void) 518 { 519 struct schedstate_percpu *spc; 520 runqueue_t *rq; 521 522 spc = &curcpu()->ci_schedstate; 523 rq = spc->spc_sched_info; 524 525 if (__predict_true((spc->spc_flags & SPCF_OFFLINE) == 0)) 526 return (global_queue.rq_bitmap | rq->rq_bitmap) != 0; 527 return rq->rq_bitmap != 0; 528 } 529 530 void 531 sched_nice(struct proc *chgp, int n) 532 { 533 534 chgp->p_nice = n; 535 (void)resetprocpriority(chgp); 536 } 537 538 /* 539 * Compute the priority of a process when running in user mode. 540 * Arrange to reschedule if the resulting priority is better 541 * than that of the current process. 542 */ 543 static void 544 resetpriority(struct lwp *l) 545 { 546 unsigned int newpriority; 547 struct proc *p = l->l_proc; 548 549 /* XXXSMP LOCK_ASSERT(mutex_owned(&p->p_stmutex)); */ 550 LOCK_ASSERT(lwp_locked(l, NULL)); 551 552 if ((l->l_flag & LW_SYSTEM) != 0) 553 return; 554 555 newpriority = PUSER + (p->p_estcpu >> ESTCPU_SHIFT) + 556 NICE_WEIGHT * (p->p_nice - NZERO); 557 newpriority = min(newpriority, MAXPRI); 558 lwp_changepri(l, newpriority); 559 } 560 561 /* 562 * Recompute priority for all LWPs in a process. 563 */ 564 static void 565 resetprocpriority(struct proc *p) 566 { 567 struct lwp *l; 568 569 KASSERT(mutex_owned(&p->p_stmutex)); 570 571 LIST_FOREACH(l, &p->p_lwps, l_sibling) { 572 lwp_lock(l); 573 resetpriority(l); 574 lwp_unlock(l); 575 } 576 } 577 578 /* 579 * We adjust the priority of the current process. The priority of a process 580 * gets worse as it accumulates CPU time. The CPU usage estimator (p_estcpu) 581 * is increased here. The formula for computing priorities (in kern_synch.c) 582 * will compute a different value each time p_estcpu increases. This can 583 * cause a switch, but unless the priority crosses a PPQ boundary the actual 584 * queue will not change. The CPU usage estimator ramps up quite quickly 585 * when the process is running (linearly), and decays away exponentially, at 586 * a rate which is proportionally slower when the system is busy. The basic 587 * principle is that the system will 90% forget that the process used a lot 588 * of CPU time in 5 * loadav seconds. This causes the system to favor 589 * processes which haven't run much recently, and to round-robin among other 590 * processes. 591 */ 592 593 void 594 sched_schedclock(struct lwp *l) 595 { 596 struct proc *p = l->l_proc; 597 598 KASSERT(!CURCPU_IDLE_P()); 599 mutex_spin_enter(&p->p_stmutex); 600 p->p_estcpu = ESTCPULIM(p->p_estcpu + (1 << ESTCPU_SHIFT)); 601 lwp_lock(l); 602 resetpriority(l); 603 mutex_spin_exit(&p->p_stmutex); 604 if ((l->l_flag & LW_SYSTEM) == 0 && l->l_priority >= PUSER) 605 l->l_priority = l->l_usrpri; 606 lwp_unlock(l); 607 } 608 609 /* 610 * sched_proc_fork: 611 * 612 * Inherit the parent's scheduler history. 613 */ 614 void 615 sched_proc_fork(struct proc *parent, struct proc *child) 616 { 617 618 KASSERT(mutex_owned(&parent->p_smutex)); 619 620 child->p_estcpu = child->p_estcpu_inherited = parent->p_estcpu; 621 child->p_forktime = sched_pstats_ticks; 622 } 623 624 /* 625 * sched_proc_exit: 626 * 627 * Chargeback parents for the sins of their children. 628 */ 629 void 630 sched_proc_exit(struct proc *parent, struct proc *child) 631 { 632 fixpt_t loadfac = loadfactor(averunnable.ldavg[0]); 633 fixpt_t estcpu; 634 635 /* XXX Only if parent != init?? */ 636 637 mutex_spin_enter(&parent->p_stmutex); 638 estcpu = decay_cpu_batch(loadfac, child->p_estcpu_inherited, 639 sched_pstats_ticks - child->p_forktime); 640 if (child->p_estcpu > estcpu) 641 parent->p_estcpu = 642 ESTCPULIM(parent->p_estcpu + child->p_estcpu - estcpu); 643 mutex_spin_exit(&parent->p_stmutex); 644 } 645 646 void 647 sched_enqueue(struct lwp *l, bool ctxswitch) 648 { 649 650 if ((l->l_flag & LW_BOUND) != 0) 651 runqueue_enqueue(l->l_cpu->ci_schedstate.spc_sched_info, l); 652 else 653 runqueue_enqueue(&global_queue, l); 654 } 655 656 /* 657 * XXXSMP When LWP dispatch (cpu_switch()) is changed to use sched_dequeue(), 658 * drop of the effective priority level from kernel to user needs to be 659 * moved here from userret(). The assignment in userret() is currently 660 * done unlocked. 661 */ 662 void 663 sched_dequeue(struct lwp *l) 664 { 665 666 if ((l->l_flag & LW_BOUND) != 0) 667 runqueue_dequeue(l->l_cpu->ci_schedstate.spc_sched_info, l); 668 else 669 runqueue_dequeue(&global_queue, l); 670 } 671 672 struct lwp * 673 sched_nextlwp(void) 674 { 675 struct schedstate_percpu *spc; 676 lwp_t *l1, *l2; 677 678 spc = &curcpu()->ci_schedstate; 679 680 /* For now, just pick the highest priority LWP. */ 681 l1 = runqueue_nextlwp(spc->spc_sched_info); 682 if (__predict_false((spc->spc_flags & SPCF_OFFLINE) != 0)) 683 return l1; 684 l2 = runqueue_nextlwp(&global_queue); 685 686 if (l1 == NULL) 687 return l2; 688 if (l2 == NULL) 689 return l1; 690 if (lwp_eprio(l2) < lwp_eprio(l1)) 691 return l2; 692 else 693 return l1; 694 } 695 696 /* Dummy */ 697 void 698 sched_lwp_fork(struct lwp *l) 699 { 700 701 } 702 703 void 704 sched_lwp_exit(struct lwp *l) 705 { 706 707 } 708 709 /* SysCtl */ 710 711 SYSCTL_SETUP(sysctl_sched_setup, "sysctl kern.sched subtree setup") 712 { 713 const struct sysctlnode *node = NULL; 714 715 sysctl_createv(clog, 0, NULL, NULL, 716 CTLFLAG_PERMANENT, 717 CTLTYPE_NODE, "kern", NULL, 718 NULL, 0, NULL, 0, 719 CTL_KERN, CTL_EOL); 720 sysctl_createv(clog, 0, NULL, &node, 721 CTLFLAG_PERMANENT, 722 CTLTYPE_NODE, "sched", 723 SYSCTL_DESCR("Scheduler options"), 724 NULL, 0, NULL, 0, 725 CTL_KERN, CTL_CREATE, CTL_EOL); 726 727 if (node != NULL) { 728 sysctl_createv(clog, 0, &node, NULL, 729 CTLFLAG_PERMANENT, 730 CTLTYPE_STRING, "name", NULL, 731 NULL, 0, __UNCONST("4.4BSD"), 0, 732 CTL_CREATE, CTL_EOL); 733 } 734 } 735 736 #if defined(DDB) 737 void 738 sched_print_runqueue(void (*pr)(const char *, ...)) 739 { 740 741 runqueue_print(&global_queue, pr); 742 } 743 #endif /* defined(DDB) */ 744