1 /*- 2 * Copyright (c) 2002-2007, Jeffrey Roberson <jeff@freebsd.org> 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice unmodified, this list of conditions, and the following 10 * disclaimer. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 16 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 17 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 18 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 19 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 20 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 21 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 22 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 23 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 24 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 25 */ 26 27 #include <sys/cdefs.h> 28 __FBSDID("$FreeBSD$"); 29 30 #include "opt_hwpmc_hooks.h" 31 #include "opt_sched.h" 32 33 #include <sys/param.h> 34 #include <sys/systm.h> 35 #include <sys/kdb.h> 36 #include <sys/kernel.h> 37 #include <sys/ktr.h> 38 #include <sys/lock.h> 39 #include <sys/mutex.h> 40 #include <sys/proc.h> 41 #include <sys/resource.h> 42 #include <sys/resourcevar.h> 43 #include <sys/sched.h> 44 #include <sys/smp.h> 45 #include <sys/sx.h> 46 #include <sys/sysctl.h> 47 #include <sys/sysproto.h> 48 #include <sys/turnstile.h> 49 #include <sys/umtx.h> 50 #include <sys/vmmeter.h> 51 #ifdef KTRACE 52 #include <sys/uio.h> 53 #include <sys/ktrace.h> 54 #endif 55 56 #ifdef HWPMC_HOOKS 57 #include <sys/pmckern.h> 58 #endif 59 60 #include <machine/cpu.h> 61 #include <machine/smp.h> 62 63 #ifndef PREEMPTION 64 #error "SCHED_ULE requires options PREEMPTION" 65 #endif 66 67 /* 68 * TODO: 69 * Pick idle from affinity group or self group first. 70 * Implement pick_score. 71 */ 72 73 /* 74 * Thread scheduler specific section. 75 */ 76 struct td_sched { 77 TAILQ_ENTRY(td_sched) ts_procq; /* (j/z) Run queue. */ 78 int ts_flags; /* (j) TSF_* flags. */ 79 struct thread *ts_thread; /* (*) Active associated thread. */ 80 u_char ts_rqindex; /* (j) Run queue index. */ 81 int ts_slptime; 82 int ts_slice; 83 struct runq *ts_runq; 84 u_char ts_cpu; /* CPU that we have affinity for. */ 85 /* The following variables are only used for pctcpu calculation */ 86 int ts_ltick; /* Last tick that we were running on */ 87 int ts_ftick; /* First tick that we were running on */ 88 int ts_ticks; /* Tick count */ 89 #ifdef SMP 90 int ts_rltick; /* Real last tick, for affinity. */ 91 #endif 92 93 /* originally from kg_sched */ 94 u_int skg_slptime; /* Number of ticks we vol. slept */ 95 u_int skg_runtime; /* Number of ticks we were running */ 96 }; 97 /* flags kept in ts_flags */ 98 #define TSF_BOUND 0x0001 /* Thread can not migrate. */ 99 #define TSF_XFERABLE 0x0002 /* Thread was added as transferable. */ 100 #define TSF_DIDRUN 0x2000 /* Thread actually ran. */ 101 102 static struct td_sched td_sched0; 103 104 /* 105 * Cpu percentage computation macros and defines. 106 * 107 * SCHED_TICK_SECS: Number of seconds to average the cpu usage across. 108 * SCHED_TICK_TARG: Number of hz ticks to average the cpu usage across. 109 * SCHED_TICK_MAX: Maximum number of ticks before scaling back. 110 * SCHED_TICK_SHIFT: Shift factor to avoid rounding away results. 111 * SCHED_TICK_HZ: Compute the number of hz ticks for a given ticks count. 112 * SCHED_TICK_TOTAL: Gives the amount of time we've been recording ticks. 113 */ 114 #define SCHED_TICK_SECS 10 115 #define SCHED_TICK_TARG (hz * SCHED_TICK_SECS) 116 #define SCHED_TICK_MAX (SCHED_TICK_TARG + hz) 117 #define SCHED_TICK_SHIFT 10 118 #define SCHED_TICK_HZ(ts) ((ts)->ts_ticks >> SCHED_TICK_SHIFT) 119 #define SCHED_TICK_TOTAL(ts) (max((ts)->ts_ltick - (ts)->ts_ftick, hz)) 120 121 /* 122 * These macros determine priorities for non-interactive threads. They are 123 * assigned a priority based on their recent cpu utilization as expressed 124 * by the ratio of ticks to the tick total. NHALF priorities at the start 125 * and end of the MIN to MAX timeshare range are only reachable with negative 126 * or positive nice respectively. 127 * 128 * PRI_RANGE: Priority range for utilization dependent priorities. 129 * PRI_NRESV: Number of nice values. 130 * PRI_TICKS: Compute a priority in PRI_RANGE from the ticks count and total. 131 * PRI_NICE: Determines the part of the priority inherited from nice. 132 */ 133 #define SCHED_PRI_NRESV (PRIO_MAX - PRIO_MIN) 134 #define SCHED_PRI_NHALF (SCHED_PRI_NRESV / 2) 135 #define SCHED_PRI_MIN (PRI_MIN_TIMESHARE + SCHED_PRI_NHALF) 136 #define SCHED_PRI_MAX (PRI_MAX_TIMESHARE - SCHED_PRI_NHALF) 137 #define SCHED_PRI_RANGE (SCHED_PRI_MAX - SCHED_PRI_MIN + 1) 138 #define SCHED_PRI_TICKS(ts) \ 139 (SCHED_TICK_HZ((ts)) / \ 140 (roundup(SCHED_TICK_TOTAL((ts)), SCHED_PRI_RANGE) / SCHED_PRI_RANGE)) 141 #define SCHED_PRI_NICE(nice) (nice) 142 143 /* 144 * These determine the interactivity of a process. Interactivity differs from 145 * cpu utilization in that it expresses the voluntary time slept vs time ran 146 * while cpu utilization includes all time not running. This more accurately 147 * models the intent of the thread. 148 * 149 * SLP_RUN_MAX: Maximum amount of sleep time + run time we'll accumulate 150 * before throttling back. 151 * SLP_RUN_FORK: Maximum slp+run time to inherit at fork time. 152 * INTERACT_MAX: Maximum interactivity value. Smaller is better. 153 * INTERACT_THRESH: Threshhold for placement on the current runq. 154 */ 155 #define SCHED_SLP_RUN_MAX ((hz * 5) << SCHED_TICK_SHIFT) 156 #define SCHED_SLP_RUN_FORK ((hz / 2) << SCHED_TICK_SHIFT) 157 #define SCHED_INTERACT_MAX (100) 158 #define SCHED_INTERACT_HALF (SCHED_INTERACT_MAX / 2) 159 #define SCHED_INTERACT_THRESH (30) 160 161 /* 162 * tickincr: Converts a stathz tick into a hz domain scaled by 163 * the shift factor. Without the shift the error rate 164 * due to rounding would be unacceptably high. 165 * realstathz: stathz is sometimes 0 and run off of hz. 166 * sched_slice: Runtime of each thread before rescheduling. 167 */ 168 static int sched_interact = SCHED_INTERACT_THRESH; 169 static int realstathz; 170 static int tickincr; 171 static int sched_slice; 172 173 /* 174 * tdq - per processor runqs and statistics. 175 */ 176 struct tdq { 177 struct runq tdq_idle; /* Queue of IDLE threads. */ 178 struct runq tdq_timeshare; /* timeshare run queue. */ 179 struct runq tdq_realtime; /* real-time run queue. */ 180 int tdq_idx; /* Current insert index. */ 181 int tdq_ridx; /* Current removal index. */ 182 int tdq_load; /* Aggregate load. */ 183 int tdq_flags; /* Thread queue flags */ 184 #ifdef SMP 185 int tdq_transferable; 186 LIST_ENTRY(tdq) tdq_siblings; /* Next in tdq group. */ 187 struct tdq_group *tdq_group; /* Our processor group. */ 188 #else 189 int tdq_sysload; /* For loadavg, !ITHD load. */ 190 #endif 191 }; 192 193 #define TDQF_BUSY 0x0001 /* Queue is marked as busy */ 194 195 #ifdef SMP 196 /* 197 * tdq groups are groups of processors which can cheaply share threads. When 198 * one processor in the group goes idle it will check the runqs of the other 199 * processors in its group prior to halting and waiting for an interrupt. 200 * These groups are suitable for SMT (Symetric Multi-Threading) and not NUMA. 201 * In a numa environment we'd want an idle bitmap per group and a two tiered 202 * load balancer. 203 */ 204 struct tdq_group { 205 int tdg_cpus; /* Count of CPUs in this tdq group. */ 206 cpumask_t tdg_cpumask; /* Mask of cpus in this group. */ 207 cpumask_t tdg_idlemask; /* Idle cpus in this group. */ 208 cpumask_t tdg_mask; /* Bit mask for first cpu. */ 209 int tdg_load; /* Total load of this group. */ 210 int tdg_transferable; /* Transferable load of this group. */ 211 LIST_HEAD(, tdq) tdg_members; /* Linked list of all members. */ 212 }; 213 214 #define SCHED_AFFINITY_DEFAULT (hz / 100) 215 #define SCHED_AFFINITY(ts) ((ts)->ts_rltick > ticks - affinity) 216 217 /* 218 * Run-time tunables. 219 */ 220 static int rebalance = 0; 221 static int pick_pri = 1; 222 static int affinity; 223 static int tryself = 1; 224 static int tryselfidle = 1; 225 static int ipi_ast = 0; 226 static int ipi_preempt = 1; 227 static int ipi_thresh = PRI_MIN_KERN; 228 static int steal_htt = 1; 229 static int steal_busy = 1; 230 static int busy_thresh = 4; 231 232 /* 233 * One thread queue per processor. 234 */ 235 static volatile cpumask_t tdq_idle; 236 static volatile cpumask_t tdq_busy; 237 static int tdg_maxid; 238 static struct tdq tdq_cpu[MAXCPU]; 239 static struct tdq_group tdq_groups[MAXCPU]; 240 static int bal_tick; 241 static int gbal_tick; 242 static int balance_groups; 243 244 #define TDQ_SELF() (&tdq_cpu[PCPU_GET(cpuid)]) 245 #define TDQ_CPU(x) (&tdq_cpu[(x)]) 246 #define TDQ_ID(x) ((x) - tdq_cpu) 247 #define TDQ_GROUP(x) (&tdq_groups[(x)]) 248 #else /* !SMP */ 249 static struct tdq tdq_cpu; 250 251 #define TDQ_SELF() (&tdq_cpu) 252 #define TDQ_CPU(x) (&tdq_cpu) 253 #endif 254 255 static void sched_priority(struct thread *); 256 static void sched_thread_priority(struct thread *, u_char); 257 static int sched_interact_score(struct thread *); 258 static void sched_interact_update(struct thread *); 259 static void sched_interact_fork(struct thread *); 260 static void sched_pctcpu_update(struct td_sched *); 261 static inline void sched_pin_td(struct thread *td); 262 static inline void sched_unpin_td(struct thread *td); 263 264 /* Operations on per processor queues */ 265 static struct td_sched * tdq_choose(struct tdq *); 266 static void tdq_setup(struct tdq *); 267 static void tdq_load_add(struct tdq *, struct td_sched *); 268 static void tdq_load_rem(struct tdq *, struct td_sched *); 269 static __inline void tdq_runq_add(struct tdq *, struct td_sched *, int); 270 static __inline void tdq_runq_rem(struct tdq *, struct td_sched *); 271 void tdq_print(int cpu); 272 static void runq_print(struct runq *rq); 273 #ifdef SMP 274 static int tdq_pickidle(struct tdq *, struct td_sched *); 275 static int tdq_pickpri(struct tdq *, struct td_sched *, int); 276 static struct td_sched *runq_steal(struct runq *); 277 static void sched_balance(void); 278 static void sched_balance_groups(void); 279 static void sched_balance_group(struct tdq_group *); 280 static void sched_balance_pair(struct tdq *, struct tdq *); 281 static void sched_smp_tick(struct thread *); 282 static void tdq_move(struct tdq *, int); 283 static int tdq_idled(struct tdq *); 284 static void tdq_notify(struct td_sched *); 285 static struct td_sched *tdq_steal(struct tdq *, int); 286 287 #define THREAD_CAN_MIGRATE(td) ((td)->td_pinned == 0) 288 #endif 289 290 static void sched_setup(void *dummy); 291 SYSINIT(sched_setup, SI_SUB_RUN_QUEUE, SI_ORDER_FIRST, sched_setup, NULL) 292 293 static void sched_initticks(void *dummy); 294 SYSINIT(sched_initticks, SI_SUB_CLOCKS, SI_ORDER_THIRD, sched_initticks, NULL) 295 296 static inline void 297 sched_pin_td(struct thread *td) 298 { 299 td->td_pinned++; 300 } 301 302 static inline void 303 sched_unpin_td(struct thread *td) 304 { 305 td->td_pinned--; 306 } 307 308 static void 309 runq_print(struct runq *rq) 310 { 311 struct rqhead *rqh; 312 struct td_sched *ts; 313 int pri; 314 int j; 315 int i; 316 317 for (i = 0; i < RQB_LEN; i++) { 318 printf("\t\trunq bits %d 0x%zx\n", 319 i, rq->rq_status.rqb_bits[i]); 320 for (j = 0; j < RQB_BPW; j++) 321 if (rq->rq_status.rqb_bits[i] & (1ul << j)) { 322 pri = j + (i << RQB_L2BPW); 323 rqh = &rq->rq_queues[pri]; 324 TAILQ_FOREACH(ts, rqh, ts_procq) { 325 printf("\t\t\ttd %p(%s) priority %d rqindex %d pri %d\n", 326 ts->ts_thread, ts->ts_thread->td_proc->p_comm, ts->ts_thread->td_priority, ts->ts_rqindex, pri); 327 } 328 } 329 } 330 } 331 332 void 333 tdq_print(int cpu) 334 { 335 struct tdq *tdq; 336 337 tdq = TDQ_CPU(cpu); 338 339 printf("tdq:\n"); 340 printf("\tload: %d\n", tdq->tdq_load); 341 printf("\ttimeshare idx: %d\n", tdq->tdq_idx); 342 printf("\ttimeshare ridx: %d\n", tdq->tdq_ridx); 343 printf("\trealtime runq:\n"); 344 runq_print(&tdq->tdq_realtime); 345 printf("\ttimeshare runq:\n"); 346 runq_print(&tdq->tdq_timeshare); 347 printf("\tidle runq:\n"); 348 runq_print(&tdq->tdq_idle); 349 #ifdef SMP 350 printf("\tload transferable: %d\n", tdq->tdq_transferable); 351 #endif 352 } 353 354 static __inline void 355 tdq_runq_add(struct tdq *tdq, struct td_sched *ts, int flags) 356 { 357 #ifdef SMP 358 if (THREAD_CAN_MIGRATE(ts->ts_thread)) { 359 tdq->tdq_transferable++; 360 tdq->tdq_group->tdg_transferable++; 361 ts->ts_flags |= TSF_XFERABLE; 362 if (tdq->tdq_transferable >= busy_thresh && 363 (tdq->tdq_flags & TDQF_BUSY) == 0) { 364 tdq->tdq_flags |= TDQF_BUSY; 365 atomic_set_int(&tdq_busy, 1 << TDQ_ID(tdq)); 366 } 367 } 368 #endif 369 if (ts->ts_runq == &tdq->tdq_timeshare) { 370 int pri; 371 372 pri = ts->ts_thread->td_priority; 373 KASSERT(pri <= PRI_MAX_TIMESHARE && pri >= PRI_MIN_TIMESHARE, 374 ("Invalid priority %d on timeshare runq", pri)); 375 /* 376 * This queue contains only priorities between MIN and MAX 377 * realtime. Use the whole queue to represent these values. 378 */ 379 #define TS_RQ_PPQ (((PRI_MAX_TIMESHARE - PRI_MIN_TIMESHARE) + 1) / RQ_NQS) 380 if ((flags & SRQ_BORROWING) == 0) { 381 pri = (pri - PRI_MIN_TIMESHARE) / TS_RQ_PPQ; 382 pri = (pri + tdq->tdq_idx) % RQ_NQS; 383 /* 384 * This effectively shortens the queue by one so we 385 * can have a one slot difference between idx and 386 * ridx while we wait for threads to drain. 387 */ 388 if (tdq->tdq_ridx != tdq->tdq_idx && 389 pri == tdq->tdq_ridx) 390 pri = (pri - 1) % RQ_NQS; 391 } else 392 pri = tdq->tdq_ridx; 393 runq_add_pri(ts->ts_runq, ts, pri, flags); 394 } else 395 runq_add(ts->ts_runq, ts, flags); 396 } 397 398 static __inline void 399 tdq_runq_rem(struct tdq *tdq, struct td_sched *ts) 400 { 401 #ifdef SMP 402 if (ts->ts_flags & TSF_XFERABLE) { 403 tdq->tdq_transferable--; 404 tdq->tdq_group->tdg_transferable--; 405 ts->ts_flags &= ~TSF_XFERABLE; 406 if (tdq->tdq_transferable < busy_thresh && 407 (tdq->tdq_flags & TDQF_BUSY)) { 408 atomic_clear_int(&tdq_busy, 1 << TDQ_ID(tdq)); 409 tdq->tdq_flags &= ~TDQF_BUSY; 410 } 411 } 412 #endif 413 if (ts->ts_runq == &tdq->tdq_timeshare) { 414 if (tdq->tdq_idx != tdq->tdq_ridx) 415 runq_remove_idx(ts->ts_runq, ts, &tdq->tdq_ridx); 416 else 417 runq_remove_idx(ts->ts_runq, ts, NULL); 418 /* 419 * For timeshare threads we update the priority here so 420 * the priority reflects the time we've been sleeping. 421 */ 422 ts->ts_ltick = ticks; 423 sched_pctcpu_update(ts); 424 sched_priority(ts->ts_thread); 425 } else 426 runq_remove(ts->ts_runq, ts); 427 } 428 429 static void 430 tdq_load_add(struct tdq *tdq, struct td_sched *ts) 431 { 432 int class; 433 mtx_assert(&sched_lock, MA_OWNED); 434 class = PRI_BASE(ts->ts_thread->td_pri_class); 435 tdq->tdq_load++; 436 CTR1(KTR_SCHED, "load: %d", tdq->tdq_load); 437 if (class != PRI_ITHD && 438 (ts->ts_thread->td_proc->p_flag & P_NOLOAD) == 0) 439 #ifdef SMP 440 tdq->tdq_group->tdg_load++; 441 #else 442 tdq->tdq_sysload++; 443 #endif 444 } 445 446 static void 447 tdq_load_rem(struct tdq *tdq, struct td_sched *ts) 448 { 449 int class; 450 mtx_assert(&sched_lock, MA_OWNED); 451 class = PRI_BASE(ts->ts_thread->td_pri_class); 452 if (class != PRI_ITHD && 453 (ts->ts_thread->td_proc->p_flag & P_NOLOAD) == 0) 454 #ifdef SMP 455 tdq->tdq_group->tdg_load--; 456 #else 457 tdq->tdq_sysload--; 458 #endif 459 tdq->tdq_load--; 460 CTR1(KTR_SCHED, "load: %d", tdq->tdq_load); 461 ts->ts_runq = NULL; 462 } 463 464 #ifdef SMP 465 static void 466 sched_smp_tick(struct thread *td) 467 { 468 struct tdq *tdq; 469 470 tdq = TDQ_SELF(); 471 if (rebalance) { 472 if (ticks >= bal_tick) 473 sched_balance(); 474 if (ticks >= gbal_tick && balance_groups) 475 sched_balance_groups(); 476 } 477 td->td_sched->ts_rltick = ticks; 478 } 479 480 /* 481 * sched_balance is a simple CPU load balancing algorithm. It operates by 482 * finding the least loaded and most loaded cpu and equalizing their load 483 * by migrating some processes. 484 * 485 * Dealing only with two CPUs at a time has two advantages. Firstly, most 486 * installations will only have 2 cpus. Secondly, load balancing too much at 487 * once can have an unpleasant effect on the system. The scheduler rarely has 488 * enough information to make perfect decisions. So this algorithm chooses 489 * algorithm simplicity and more gradual effects on load in larger systems. 490 * 491 * It could be improved by considering the priorities and slices assigned to 492 * each task prior to balancing them. There are many pathological cases with 493 * any approach and so the semi random algorithm below may work as well as any. 494 * 495 */ 496 static void 497 sched_balance(void) 498 { 499 struct tdq_group *high; 500 struct tdq_group *low; 501 struct tdq_group *tdg; 502 int cnt; 503 int i; 504 505 bal_tick = ticks + (random() % (hz * 2)); 506 if (smp_started == 0) 507 return; 508 low = high = NULL; 509 i = random() % (tdg_maxid + 1); 510 for (cnt = 0; cnt <= tdg_maxid; cnt++) { 511 tdg = TDQ_GROUP(i); 512 /* 513 * Find the CPU with the highest load that has some 514 * threads to transfer. 515 */ 516 if ((high == NULL || tdg->tdg_load > high->tdg_load) 517 && tdg->tdg_transferable) 518 high = tdg; 519 if (low == NULL || tdg->tdg_load < low->tdg_load) 520 low = tdg; 521 if (++i > tdg_maxid) 522 i = 0; 523 } 524 if (low != NULL && high != NULL && high != low) 525 sched_balance_pair(LIST_FIRST(&high->tdg_members), 526 LIST_FIRST(&low->tdg_members)); 527 } 528 529 static void 530 sched_balance_groups(void) 531 { 532 int i; 533 534 gbal_tick = ticks + (random() % (hz * 2)); 535 mtx_assert(&sched_lock, MA_OWNED); 536 if (smp_started) 537 for (i = 0; i <= tdg_maxid; i++) 538 sched_balance_group(TDQ_GROUP(i)); 539 } 540 541 static void 542 sched_balance_group(struct tdq_group *tdg) 543 { 544 struct tdq *tdq; 545 struct tdq *high; 546 struct tdq *low; 547 int load; 548 549 if (tdg->tdg_transferable == 0) 550 return; 551 low = NULL; 552 high = NULL; 553 LIST_FOREACH(tdq, &tdg->tdg_members, tdq_siblings) { 554 load = tdq->tdq_load; 555 if (high == NULL || load > high->tdq_load) 556 high = tdq; 557 if (low == NULL || load < low->tdq_load) 558 low = tdq; 559 } 560 if (high != NULL && low != NULL && high != low) 561 sched_balance_pair(high, low); 562 } 563 564 static void 565 sched_balance_pair(struct tdq *high, struct tdq *low) 566 { 567 int transferable; 568 int high_load; 569 int low_load; 570 int move; 571 int diff; 572 int i; 573 574 /* 575 * If we're transfering within a group we have to use this specific 576 * tdq's transferable count, otherwise we can steal from other members 577 * of the group. 578 */ 579 if (high->tdq_group == low->tdq_group) { 580 transferable = high->tdq_transferable; 581 high_load = high->tdq_load; 582 low_load = low->tdq_load; 583 } else { 584 transferable = high->tdq_group->tdg_transferable; 585 high_load = high->tdq_group->tdg_load; 586 low_load = low->tdq_group->tdg_load; 587 } 588 if (transferable == 0) 589 return; 590 /* 591 * Determine what the imbalance is and then adjust that to how many 592 * threads we actually have to give up (transferable). 593 */ 594 diff = high_load - low_load; 595 move = diff / 2; 596 if (diff & 0x1) 597 move++; 598 move = min(move, transferable); 599 for (i = 0; i < move; i++) 600 tdq_move(high, TDQ_ID(low)); 601 return; 602 } 603 604 static void 605 tdq_move(struct tdq *from, int cpu) 606 { 607 struct tdq *tdq; 608 struct tdq *to; 609 struct td_sched *ts; 610 611 tdq = from; 612 to = TDQ_CPU(cpu); 613 ts = tdq_steal(tdq, 1); 614 if (ts == NULL) { 615 struct tdq_group *tdg; 616 617 tdg = tdq->tdq_group; 618 LIST_FOREACH(tdq, &tdg->tdg_members, tdq_siblings) { 619 if (tdq == from || tdq->tdq_transferable == 0) 620 continue; 621 ts = tdq_steal(tdq, 1); 622 break; 623 } 624 if (ts == NULL) 625 panic("tdq_move: No threads available with a " 626 "transferable count of %d\n", 627 tdg->tdg_transferable); 628 } 629 if (tdq == to) 630 return; 631 sched_rem(ts->ts_thread); 632 ts->ts_cpu = cpu; 633 sched_pin_td(ts->ts_thread); 634 sched_add(ts->ts_thread, SRQ_YIELDING); 635 sched_unpin_td(ts->ts_thread); 636 } 637 638 static int 639 tdq_idled(struct tdq *tdq) 640 { 641 struct tdq_group *tdg; 642 struct tdq *steal; 643 struct td_sched *ts; 644 645 tdg = tdq->tdq_group; 646 /* 647 * If we're in a cpu group, try and steal threads from another cpu in 648 * the group before idling. 649 */ 650 if (steal_htt && tdg->tdg_cpus > 1 && tdg->tdg_transferable) { 651 LIST_FOREACH(steal, &tdg->tdg_members, tdq_siblings) { 652 if (steal == tdq || steal->tdq_transferable == 0) 653 continue; 654 ts = tdq_steal(steal, 0); 655 if (ts) 656 goto steal; 657 } 658 } 659 if (steal_busy) { 660 while (tdq_busy) { 661 int cpu; 662 663 cpu = ffs(tdq_busy); 664 if (cpu == 0) 665 break; 666 cpu--; 667 steal = TDQ_CPU(cpu); 668 if (steal->tdq_transferable == 0) 669 continue; 670 ts = tdq_steal(steal, 1); 671 if (ts == NULL) 672 continue; 673 CTR5(KTR_SCHED, 674 "tdq_idled: stealing td %p(%s) pri %d from %d busy 0x%X", 675 ts->ts_thread, ts->ts_thread->td_proc->p_comm, 676 ts->ts_thread->td_priority, cpu, tdq_busy); 677 goto steal; 678 } 679 } 680 /* 681 * We only set the idled bit when all of the cpus in the group are 682 * idle. Otherwise we could get into a situation where a thread bounces 683 * back and forth between two idle cores on seperate physical CPUs. 684 */ 685 tdg->tdg_idlemask |= PCPU_GET(cpumask); 686 if (tdg->tdg_idlemask == tdg->tdg_cpumask) 687 atomic_set_int(&tdq_idle, tdg->tdg_mask); 688 return (1); 689 steal: 690 sched_rem(ts->ts_thread); 691 ts->ts_cpu = PCPU_GET(cpuid); 692 sched_pin_td(ts->ts_thread); 693 sched_add(ts->ts_thread, SRQ_YIELDING); 694 sched_unpin_td(ts->ts_thread); 695 696 return (0); 697 } 698 699 static void 700 tdq_notify(struct td_sched *ts) 701 { 702 struct thread *td; 703 struct pcpu *pcpu; 704 int prio; 705 int cpu; 706 707 prio = ts->ts_thread->td_priority; 708 cpu = ts->ts_cpu; 709 pcpu = pcpu_find(cpu); 710 td = pcpu->pc_curthread; 711 712 /* 713 * If our priority is not better than the current priority there is 714 * nothing to do. 715 */ 716 if (prio > td->td_priority) 717 return; 718 /* Always set NEEDRESCHED. */ 719 td->td_flags |= TDF_NEEDRESCHED; 720 /* 721 * IPI if we exceed the threshold or if the target cpu is running an 722 * idle thread. 723 */ 724 if (prio > ipi_thresh && td->td_priority < PRI_MIN_IDLE) 725 return; 726 if (ipi_ast) 727 ipi_selected(1 << cpu, IPI_AST); 728 else if (ipi_preempt) 729 ipi_selected(1 << cpu, IPI_PREEMPT); 730 } 731 732 static struct td_sched * 733 runq_steal(struct runq *rq) 734 { 735 struct rqhead *rqh; 736 struct rqbits *rqb; 737 struct td_sched *ts; 738 int word; 739 int bit; 740 741 mtx_assert(&sched_lock, MA_OWNED); 742 rqb = &rq->rq_status; 743 for (word = 0; word < RQB_LEN; word++) { 744 if (rqb->rqb_bits[word] == 0) 745 continue; 746 for (bit = 0; bit < RQB_BPW; bit++) { 747 if ((rqb->rqb_bits[word] & (1ul << bit)) == 0) 748 continue; 749 rqh = &rq->rq_queues[bit + (word << RQB_L2BPW)]; 750 TAILQ_FOREACH(ts, rqh, ts_procq) { 751 if (THREAD_CAN_MIGRATE(ts->ts_thread)) 752 return (ts); 753 } 754 } 755 } 756 return (NULL); 757 } 758 759 static struct td_sched * 760 tdq_steal(struct tdq *tdq, int stealidle) 761 { 762 struct td_sched *ts; 763 764 /* 765 * Steal from next first to try to get a non-interactive task that 766 * may not have run for a while. 767 * XXX Need to effect steal order for timeshare threads. 768 */ 769 if ((ts = runq_steal(&tdq->tdq_realtime)) != NULL) 770 return (ts); 771 if ((ts = runq_steal(&tdq->tdq_timeshare)) != NULL) 772 return (ts); 773 if (stealidle) 774 return (runq_steal(&tdq->tdq_idle)); 775 return (NULL); 776 } 777 778 int 779 tdq_pickidle(struct tdq *tdq, struct td_sched *ts) 780 { 781 struct tdq_group *tdg; 782 int self; 783 int cpu; 784 785 self = PCPU_GET(cpuid); 786 if (smp_started == 0) 787 return (self); 788 /* 789 * If the current CPU has idled, just run it here. 790 */ 791 if ((tdq->tdq_group->tdg_idlemask & PCPU_GET(cpumask)) != 0) 792 return (self); 793 /* 794 * Try the last group we ran on. 795 */ 796 tdg = TDQ_CPU(ts->ts_cpu)->tdq_group; 797 cpu = ffs(tdg->tdg_idlemask); 798 if (cpu) 799 return (cpu - 1); 800 /* 801 * Search for an idle group. 802 */ 803 cpu = ffs(tdq_idle); 804 if (cpu) 805 return (cpu - 1); 806 /* 807 * XXX If there are no idle groups, check for an idle core. 808 */ 809 /* 810 * No idle CPUs? 811 */ 812 return (self); 813 } 814 815 static int 816 tdq_pickpri(struct tdq *tdq, struct td_sched *ts, int flags) 817 { 818 struct pcpu *pcpu; 819 int lowpri; 820 int lowcpu; 821 int lowload; 822 int load; 823 int self; 824 int pri; 825 int cpu; 826 827 self = PCPU_GET(cpuid); 828 if (smp_started == 0) 829 return (self); 830 831 pri = ts->ts_thread->td_priority; 832 /* 833 * Regardless of affinity, if the last cpu is idle send it there. 834 */ 835 pcpu = pcpu_find(ts->ts_cpu); 836 if (pcpu->pc_curthread->td_priority > PRI_MIN_IDLE) { 837 CTR5(KTR_SCHED, 838 "ts_cpu %d idle, ltick %d ticks %d pri %d curthread %d", 839 ts->ts_cpu, ts->ts_rltick, ticks, pri, 840 pcpu->pc_curthread->td_priority); 841 return (ts->ts_cpu); 842 } 843 /* 844 * If we have affinity, try to place it on the cpu we last ran on. 845 */ 846 if (SCHED_AFFINITY(ts) && pcpu->pc_curthread->td_priority > pri) { 847 CTR5(KTR_SCHED, 848 "affinity for %d, ltick %d ticks %d pri %d curthread %d", 849 ts->ts_cpu, ts->ts_rltick, ticks, pri, 850 pcpu->pc_curthread->td_priority); 851 return (ts->ts_cpu); 852 } 853 /* 854 * Try ourself first; If we're running something lower priority this 855 * may have some locality with the waking thread and execute faster 856 * here. 857 */ 858 if (tryself) { 859 /* 860 * If we're being awoken by an interrupt thread or the waker 861 * is going right to sleep run here as well. 862 */ 863 if ((TDQ_SELF()->tdq_load == 1) && (flags & SRQ_YIELDING || 864 curthread->td_pri_class == PRI_ITHD)) { 865 CTR2(KTR_SCHED, "tryself load %d flags %d", 866 TDQ_SELF()->tdq_load, flags); 867 return (self); 868 } 869 } 870 /* 871 * Look for an idle group. 872 */ 873 CTR1(KTR_SCHED, "tdq_idle %X", tdq_idle); 874 cpu = ffs(tdq_idle); 875 if (cpu) 876 return (cpu - 1); 877 if (tryselfidle && pri < curthread->td_priority) { 878 CTR1(KTR_SCHED, "tryself %d", 879 curthread->td_priority); 880 return (self); 881 } 882 /* 883 * Now search for the cpu running the lowest priority thread with 884 * the least load. 885 */ 886 lowload = 0; 887 lowpri = lowcpu = 0; 888 for (cpu = 0; cpu <= mp_maxid; cpu++) { 889 if (CPU_ABSENT(cpu)) 890 continue; 891 pcpu = pcpu_find(cpu); 892 pri = pcpu->pc_curthread->td_priority; 893 CTR4(KTR_SCHED, 894 "cpu %d pri %d lowcpu %d lowpri %d", 895 cpu, pri, lowcpu, lowpri); 896 if (pri < lowpri) 897 continue; 898 load = TDQ_CPU(cpu)->tdq_load; 899 if (lowpri && lowpri == pri && load > lowload) 900 continue; 901 lowpri = pri; 902 lowcpu = cpu; 903 lowload = load; 904 } 905 906 return (lowcpu); 907 } 908 909 #endif /* SMP */ 910 911 /* 912 * Pick the highest priority task we have and return it. 913 */ 914 915 static struct td_sched * 916 tdq_choose(struct tdq *tdq) 917 { 918 struct td_sched *ts; 919 920 mtx_assert(&sched_lock, MA_OWNED); 921 922 ts = runq_choose(&tdq->tdq_realtime); 923 if (ts != NULL) { 924 KASSERT(ts->ts_thread->td_priority <= PRI_MAX_REALTIME, 925 ("tdq_choose: Invalid priority on realtime queue %d", 926 ts->ts_thread->td_priority)); 927 return (ts); 928 } 929 ts = runq_choose_from(&tdq->tdq_timeshare, tdq->tdq_ridx); 930 if (ts != NULL) { 931 KASSERT(ts->ts_thread->td_priority <= PRI_MAX_TIMESHARE && 932 ts->ts_thread->td_priority >= PRI_MIN_TIMESHARE, 933 ("tdq_choose: Invalid priority on timeshare queue %d", 934 ts->ts_thread->td_priority)); 935 return (ts); 936 } 937 938 ts = runq_choose(&tdq->tdq_idle); 939 if (ts != NULL) { 940 KASSERT(ts->ts_thread->td_priority >= PRI_MIN_IDLE, 941 ("tdq_choose: Invalid priority on idle queue %d", 942 ts->ts_thread->td_priority)); 943 return (ts); 944 } 945 946 return (NULL); 947 } 948 949 static void 950 tdq_setup(struct tdq *tdq) 951 { 952 runq_init(&tdq->tdq_realtime); 953 runq_init(&tdq->tdq_timeshare); 954 runq_init(&tdq->tdq_idle); 955 tdq->tdq_load = 0; 956 } 957 958 static void 959 sched_setup(void *dummy) 960 { 961 #ifdef SMP 962 int i; 963 #endif 964 965 /* 966 * To avoid divide-by-zero, we set realstathz a dummy value 967 * in case which sched_clock() called before sched_initticks(). 968 */ 969 realstathz = hz; 970 sched_slice = (realstathz/7); /* 140ms */ 971 tickincr = 1 << SCHED_TICK_SHIFT; 972 973 #ifdef SMP 974 balance_groups = 0; 975 /* 976 * Initialize the tdqs. 977 */ 978 for (i = 0; i < MAXCPU; i++) { 979 struct tdq *tdq; 980 981 tdq = &tdq_cpu[i]; 982 tdq_setup(&tdq_cpu[i]); 983 } 984 if (smp_topology == NULL) { 985 struct tdq_group *tdg; 986 struct tdq *tdq; 987 int cpus; 988 989 for (cpus = 0, i = 0; i < MAXCPU; i++) { 990 if (CPU_ABSENT(i)) 991 continue; 992 tdq = &tdq_cpu[i]; 993 tdg = &tdq_groups[cpus]; 994 /* 995 * Setup a tdq group with one member. 996 */ 997 tdq->tdq_transferable = 0; 998 tdq->tdq_group = tdg; 999 tdg->tdg_cpus = 1; 1000 tdg->tdg_idlemask = 0; 1001 tdg->tdg_cpumask = tdg->tdg_mask = 1 << i; 1002 tdg->tdg_load = 0; 1003 tdg->tdg_transferable = 0; 1004 LIST_INIT(&tdg->tdg_members); 1005 LIST_INSERT_HEAD(&tdg->tdg_members, tdq, tdq_siblings); 1006 cpus++; 1007 } 1008 tdg_maxid = cpus - 1; 1009 } else { 1010 struct tdq_group *tdg; 1011 struct cpu_group *cg; 1012 int j; 1013 1014 for (i = 0; i < smp_topology->ct_count; i++) { 1015 cg = &smp_topology->ct_group[i]; 1016 tdg = &tdq_groups[i]; 1017 /* 1018 * Initialize the group. 1019 */ 1020 tdg->tdg_idlemask = 0; 1021 tdg->tdg_load = 0; 1022 tdg->tdg_transferable = 0; 1023 tdg->tdg_cpus = cg->cg_count; 1024 tdg->tdg_cpumask = cg->cg_mask; 1025 LIST_INIT(&tdg->tdg_members); 1026 /* 1027 * Find all of the group members and add them. 1028 */ 1029 for (j = 0; j < MAXCPU; j++) { 1030 if ((cg->cg_mask & (1 << j)) != 0) { 1031 if (tdg->tdg_mask == 0) 1032 tdg->tdg_mask = 1 << j; 1033 tdq_cpu[j].tdq_transferable = 0; 1034 tdq_cpu[j].tdq_group = tdg; 1035 LIST_INSERT_HEAD(&tdg->tdg_members, 1036 &tdq_cpu[j], tdq_siblings); 1037 } 1038 } 1039 if (tdg->tdg_cpus > 1) 1040 balance_groups = 1; 1041 } 1042 tdg_maxid = smp_topology->ct_count - 1; 1043 } 1044 /* 1045 * Stagger the group and global load balancer so they do not 1046 * interfere with each other. 1047 */ 1048 bal_tick = ticks + hz; 1049 if (balance_groups) 1050 gbal_tick = ticks + (hz / 2); 1051 #else 1052 tdq_setup(TDQ_SELF()); 1053 #endif 1054 mtx_lock_spin(&sched_lock); 1055 tdq_load_add(TDQ_SELF(), &td_sched0); 1056 mtx_unlock_spin(&sched_lock); 1057 } 1058 1059 /* ARGSUSED */ 1060 static void 1061 sched_initticks(void *dummy) 1062 { 1063 mtx_lock_spin(&sched_lock); 1064 realstathz = stathz ? stathz : hz; 1065 sched_slice = (realstathz/7); /* ~140ms */ 1066 1067 /* 1068 * tickincr is shifted out by 10 to avoid rounding errors due to 1069 * hz not being evenly divisible by stathz on all platforms. 1070 */ 1071 tickincr = (hz << SCHED_TICK_SHIFT) / realstathz; 1072 /* 1073 * This does not work for values of stathz that are more than 1074 * 1 << SCHED_TICK_SHIFT * hz. In practice this does not happen. 1075 */ 1076 if (tickincr == 0) 1077 tickincr = 1; 1078 #ifdef SMP 1079 affinity = SCHED_AFFINITY_DEFAULT; 1080 #endif 1081 mtx_unlock_spin(&sched_lock); 1082 } 1083 1084 1085 /* 1086 * Scale the scheduling priority according to the "interactivity" of this 1087 * process. 1088 */ 1089 static void 1090 sched_priority(struct thread *td) 1091 { 1092 int score; 1093 int pri; 1094 1095 if (td->td_pri_class != PRI_TIMESHARE) 1096 return; 1097 /* 1098 * If the score is interactive we place the thread in the realtime 1099 * queue with a priority that is less than kernel and interrupt 1100 * priorities. These threads are not subject to nice restrictions. 1101 * 1102 * Scores greater than this are placed on the normal realtime queue 1103 * where the priority is partially decided by the most recent cpu 1104 * utilization and the rest is decided by nice value. 1105 */ 1106 score = sched_interact_score(td); 1107 if (score < sched_interact) { 1108 pri = PRI_MIN_REALTIME; 1109 pri += ((PRI_MAX_REALTIME - PRI_MIN_REALTIME) / sched_interact) 1110 * score; 1111 KASSERT(pri >= PRI_MIN_REALTIME && pri <= PRI_MAX_REALTIME, 1112 ("sched_priority: invalid interactive priority %d score %d", 1113 pri, score)); 1114 } else { 1115 pri = SCHED_PRI_MIN; 1116 if (td->td_sched->ts_ticks) 1117 pri += SCHED_PRI_TICKS(td->td_sched); 1118 pri += SCHED_PRI_NICE(td->td_proc->p_nice); 1119 if (!(pri >= PRI_MIN_TIMESHARE && pri <= PRI_MAX_TIMESHARE)) { 1120 static int once = 1; 1121 if (once) { 1122 printf("sched_priority: invalid priority %d", 1123 pri); 1124 printf("nice %d, ticks %d ftick %d ltick %d tick pri %d\n", 1125 td->td_proc->p_nice, 1126 td->td_sched->ts_ticks, 1127 td->td_sched->ts_ftick, 1128 td->td_sched->ts_ltick, 1129 SCHED_PRI_TICKS(td->td_sched)); 1130 once = 0; 1131 } 1132 pri = min(max(pri, PRI_MIN_TIMESHARE), 1133 PRI_MAX_TIMESHARE); 1134 } 1135 } 1136 sched_user_prio(td, pri); 1137 1138 return; 1139 } 1140 1141 /* 1142 * This routine enforces a maximum limit on the amount of scheduling history 1143 * kept. It is called after either the slptime or runtime is adjusted. 1144 */ 1145 static void 1146 sched_interact_update(struct thread *td) 1147 { 1148 struct td_sched *ts; 1149 u_int sum; 1150 1151 ts = td->td_sched; 1152 sum = ts->skg_runtime + ts->skg_slptime; 1153 if (sum < SCHED_SLP_RUN_MAX) 1154 return; 1155 /* 1156 * This only happens from two places: 1157 * 1) We have added an unusual amount of run time from fork_exit. 1158 * 2) We have added an unusual amount of sleep time from sched_sleep(). 1159 */ 1160 if (sum > SCHED_SLP_RUN_MAX * 2) { 1161 if (ts->skg_runtime > ts->skg_slptime) { 1162 ts->skg_runtime = SCHED_SLP_RUN_MAX; 1163 ts->skg_slptime = 1; 1164 } else { 1165 ts->skg_slptime = SCHED_SLP_RUN_MAX; 1166 ts->skg_runtime = 1; 1167 } 1168 return; 1169 } 1170 /* 1171 * If we have exceeded by more than 1/5th then the algorithm below 1172 * will not bring us back into range. Dividing by two here forces 1173 * us into the range of [4/5 * SCHED_INTERACT_MAX, SCHED_INTERACT_MAX] 1174 */ 1175 if (sum > (SCHED_SLP_RUN_MAX / 5) * 6) { 1176 ts->skg_runtime /= 2; 1177 ts->skg_slptime /= 2; 1178 return; 1179 } 1180 ts->skg_runtime = (ts->skg_runtime / 5) * 4; 1181 ts->skg_slptime = (ts->skg_slptime / 5) * 4; 1182 } 1183 1184 static void 1185 sched_interact_fork(struct thread *td) 1186 { 1187 int ratio; 1188 int sum; 1189 1190 sum = td->td_sched->skg_runtime + td->td_sched->skg_slptime; 1191 if (sum > SCHED_SLP_RUN_FORK) { 1192 ratio = sum / SCHED_SLP_RUN_FORK; 1193 td->td_sched->skg_runtime /= ratio; 1194 td->td_sched->skg_slptime /= ratio; 1195 } 1196 } 1197 1198 static int 1199 sched_interact_score(struct thread *td) 1200 { 1201 int div; 1202 1203 if (td->td_sched->skg_runtime > td->td_sched->skg_slptime) { 1204 div = max(1, td->td_sched->skg_runtime / SCHED_INTERACT_HALF); 1205 return (SCHED_INTERACT_HALF + 1206 (SCHED_INTERACT_HALF - (td->td_sched->skg_slptime / div))); 1207 } if (td->td_sched->skg_slptime > td->td_sched->skg_runtime) { 1208 div = max(1, td->td_sched->skg_slptime / SCHED_INTERACT_HALF); 1209 return (td->td_sched->skg_runtime / div); 1210 } 1211 1212 /* 1213 * This can happen if slptime and runtime are 0. 1214 */ 1215 return (0); 1216 1217 } 1218 1219 /* 1220 * Called from proc0_init() to bootstrap the scheduler. 1221 */ 1222 void 1223 schedinit(void) 1224 { 1225 1226 /* 1227 * Set up the scheduler specific parts of proc0. 1228 */ 1229 proc0.p_sched = NULL; /* XXX */ 1230 thread0.td_sched = &td_sched0; 1231 td_sched0.ts_ltick = ticks; 1232 td_sched0.ts_ftick = ticks; 1233 td_sched0.ts_thread = &thread0; 1234 } 1235 1236 /* 1237 * This is only somewhat accurate since given many processes of the same 1238 * priority they will switch when their slices run out, which will be 1239 * at most sched_slice stathz ticks. 1240 */ 1241 int 1242 sched_rr_interval(void) 1243 { 1244 1245 /* Convert sched_slice to hz */ 1246 return (hz/(realstathz/sched_slice)); 1247 } 1248 1249 static void 1250 sched_pctcpu_update(struct td_sched *ts) 1251 { 1252 1253 if (ts->ts_ticks == 0) 1254 return; 1255 if (ticks - (hz / 10) < ts->ts_ltick && 1256 SCHED_TICK_TOTAL(ts) < SCHED_TICK_MAX) 1257 return; 1258 /* 1259 * Adjust counters and watermark for pctcpu calc. 1260 */ 1261 if (ts->ts_ltick > ticks - SCHED_TICK_TARG) 1262 ts->ts_ticks = (ts->ts_ticks / (ticks - ts->ts_ftick)) * 1263 SCHED_TICK_TARG; 1264 else 1265 ts->ts_ticks = 0; 1266 ts->ts_ltick = ticks; 1267 ts->ts_ftick = ts->ts_ltick - SCHED_TICK_TARG; 1268 } 1269 1270 static void 1271 sched_thread_priority(struct thread *td, u_char prio) 1272 { 1273 struct td_sched *ts; 1274 1275 CTR6(KTR_SCHED, "sched_prio: %p(%s) prio %d newprio %d by %p(%s)", 1276 td, td->td_proc->p_comm, td->td_priority, prio, curthread, 1277 curthread->td_proc->p_comm); 1278 ts = td->td_sched; 1279 mtx_assert(&sched_lock, MA_OWNED); 1280 if (td->td_priority == prio) 1281 return; 1282 1283 if (TD_ON_RUNQ(td) && prio < td->td_priority) { 1284 /* 1285 * If the priority has been elevated due to priority 1286 * propagation, we may have to move ourselves to a new 1287 * queue. This could be optimized to not re-add in some 1288 * cases. 1289 */ 1290 sched_rem(td); 1291 td->td_priority = prio; 1292 sched_add(td, SRQ_BORROWING); 1293 } else 1294 td->td_priority = prio; 1295 } 1296 1297 /* 1298 * Update a thread's priority when it is lent another thread's 1299 * priority. 1300 */ 1301 void 1302 sched_lend_prio(struct thread *td, u_char prio) 1303 { 1304 1305 td->td_flags |= TDF_BORROWING; 1306 sched_thread_priority(td, prio); 1307 } 1308 1309 /* 1310 * Restore a thread's priority when priority propagation is 1311 * over. The prio argument is the minimum priority the thread 1312 * needs to have to satisfy other possible priority lending 1313 * requests. If the thread's regular priority is less 1314 * important than prio, the thread will keep a priority boost 1315 * of prio. 1316 */ 1317 void 1318 sched_unlend_prio(struct thread *td, u_char prio) 1319 { 1320 u_char base_pri; 1321 1322 if (td->td_base_pri >= PRI_MIN_TIMESHARE && 1323 td->td_base_pri <= PRI_MAX_TIMESHARE) 1324 base_pri = td->td_user_pri; 1325 else 1326 base_pri = td->td_base_pri; 1327 if (prio >= base_pri) { 1328 td->td_flags &= ~TDF_BORROWING; 1329 sched_thread_priority(td, base_pri); 1330 } else 1331 sched_lend_prio(td, prio); 1332 } 1333 1334 void 1335 sched_prio(struct thread *td, u_char prio) 1336 { 1337 u_char oldprio; 1338 1339 /* First, update the base priority. */ 1340 td->td_base_pri = prio; 1341 1342 /* 1343 * If the thread is borrowing another thread's priority, don't 1344 * ever lower the priority. 1345 */ 1346 if (td->td_flags & TDF_BORROWING && td->td_priority < prio) 1347 return; 1348 1349 /* Change the real priority. */ 1350 oldprio = td->td_priority; 1351 sched_thread_priority(td, prio); 1352 1353 /* 1354 * If the thread is on a turnstile, then let the turnstile update 1355 * its state. 1356 */ 1357 if (TD_ON_LOCK(td) && oldprio != prio) 1358 turnstile_adjust(td, oldprio); 1359 } 1360 1361 void 1362 sched_user_prio(struct thread *td, u_char prio) 1363 { 1364 u_char oldprio; 1365 1366 td->td_base_user_pri = prio; 1367 if (td->td_flags & TDF_UBORROWING && td->td_user_pri <= prio) 1368 return; 1369 oldprio = td->td_user_pri; 1370 td->td_user_pri = prio; 1371 1372 if (TD_ON_UPILOCK(td) && oldprio != prio) 1373 umtx_pi_adjust(td, oldprio); 1374 } 1375 1376 void 1377 sched_lend_user_prio(struct thread *td, u_char prio) 1378 { 1379 u_char oldprio; 1380 1381 td->td_flags |= TDF_UBORROWING; 1382 1383 oldprio = td->td_user_pri; 1384 td->td_user_pri = prio; 1385 1386 if (TD_ON_UPILOCK(td) && oldprio != prio) 1387 umtx_pi_adjust(td, oldprio); 1388 } 1389 1390 void 1391 sched_unlend_user_prio(struct thread *td, u_char prio) 1392 { 1393 u_char base_pri; 1394 1395 base_pri = td->td_base_user_pri; 1396 if (prio >= base_pri) { 1397 td->td_flags &= ~TDF_UBORROWING; 1398 sched_user_prio(td, base_pri); 1399 } else 1400 sched_lend_user_prio(td, prio); 1401 } 1402 1403 void 1404 sched_switch(struct thread *td, struct thread *newtd, int flags) 1405 { 1406 struct tdq *tdq; 1407 struct td_sched *ts; 1408 int preempt; 1409 1410 mtx_assert(&sched_lock, MA_OWNED); 1411 1412 preempt = flags & SW_PREEMPT; 1413 tdq = TDQ_SELF(); 1414 ts = td->td_sched; 1415 td->td_lastcpu = td->td_oncpu; 1416 td->td_oncpu = NOCPU; 1417 td->td_flags &= ~TDF_NEEDRESCHED; 1418 td->td_owepreempt = 0; 1419 /* 1420 * If the thread has been assigned it may be in the process of switching 1421 * to the new cpu. This is the case in sched_bind(). 1422 */ 1423 if (td == PCPU_GET(idlethread)) { 1424 TD_SET_CAN_RUN(td); 1425 } else { 1426 tdq_load_rem(tdq, ts); 1427 if (TD_IS_RUNNING(td)) { 1428 /* 1429 * Don't allow the thread to migrate 1430 * from a preemption. 1431 */ 1432 if (preempt) 1433 sched_pin_td(td); 1434 sched_add(td, preempt ? 1435 SRQ_OURSELF|SRQ_YIELDING|SRQ_PREEMPTED : 1436 SRQ_OURSELF|SRQ_YIELDING); 1437 if (preempt) 1438 sched_unpin_td(td); 1439 } 1440 } 1441 if (newtd != NULL) { 1442 /* 1443 * If we bring in a thread account for it as if it had been 1444 * added to the run queue and then chosen. 1445 */ 1446 newtd->td_sched->ts_flags |= TSF_DIDRUN; 1447 TD_SET_RUNNING(newtd); 1448 tdq_load_add(TDQ_SELF(), newtd->td_sched); 1449 } else 1450 newtd = choosethread(); 1451 if (td != newtd) { 1452 #ifdef HWPMC_HOOKS 1453 if (PMC_PROC_IS_USING_PMCS(td->td_proc)) 1454 PMC_SWITCH_CONTEXT(td, PMC_FN_CSW_OUT); 1455 #endif 1456 1457 cpu_switch(td, newtd); 1458 #ifdef HWPMC_HOOKS 1459 if (PMC_PROC_IS_USING_PMCS(td->td_proc)) 1460 PMC_SWITCH_CONTEXT(td, PMC_FN_CSW_IN); 1461 #endif 1462 } 1463 sched_lock.mtx_lock = (uintptr_t)td; 1464 td->td_oncpu = PCPU_GET(cpuid); 1465 } 1466 1467 void 1468 sched_nice(struct proc *p, int nice) 1469 { 1470 struct thread *td; 1471 1472 PROC_LOCK_ASSERT(p, MA_OWNED); 1473 mtx_assert(&sched_lock, MA_OWNED); 1474 1475 p->p_nice = nice; 1476 FOREACH_THREAD_IN_PROC(p, td) { 1477 sched_priority(td); 1478 sched_prio(td, td->td_base_user_pri); 1479 } 1480 } 1481 1482 void 1483 sched_sleep(struct thread *td) 1484 { 1485 1486 mtx_assert(&sched_lock, MA_OWNED); 1487 1488 td->td_sched->ts_slptime = ticks; 1489 } 1490 1491 void 1492 sched_wakeup(struct thread *td) 1493 { 1494 int slptime; 1495 1496 mtx_assert(&sched_lock, MA_OWNED); 1497 1498 /* 1499 * If we slept for more than a tick update our interactivity and 1500 * priority. 1501 */ 1502 slptime = td->td_sched->ts_slptime; 1503 td->td_sched->ts_slptime = 0; 1504 if (slptime && slptime != ticks) { 1505 u_int hzticks; 1506 1507 hzticks = (ticks - slptime) << SCHED_TICK_SHIFT; 1508 td->td_sched->skg_slptime += hzticks; 1509 sched_interact_update(td); 1510 sched_pctcpu_update(td->td_sched); 1511 sched_priority(td); 1512 } 1513 sched_add(td, SRQ_BORING); 1514 } 1515 1516 /* 1517 * Penalize the parent for creating a new child and initialize the child's 1518 * priority. 1519 */ 1520 void 1521 sched_fork(struct thread *td, struct thread *child) 1522 { 1523 mtx_assert(&sched_lock, MA_OWNED); 1524 sched_fork_thread(td, child); 1525 /* 1526 * Penalize the parent and child for forking. 1527 */ 1528 sched_interact_fork(child); 1529 sched_priority(child); 1530 td->td_sched->skg_runtime += tickincr; 1531 sched_interact_update(td); 1532 sched_priority(td); 1533 } 1534 1535 void 1536 sched_fork_thread(struct thread *td, struct thread *child) 1537 { 1538 struct td_sched *ts; 1539 struct td_sched *ts2; 1540 1541 /* 1542 * Initialize child. 1543 */ 1544 sched_newthread(child); 1545 ts = td->td_sched; 1546 ts2 = child->td_sched; 1547 ts2->ts_cpu = ts->ts_cpu; 1548 ts2->ts_runq = NULL; 1549 /* 1550 * Grab our parents cpu estimation information and priority. 1551 */ 1552 ts2->ts_ticks = ts->ts_ticks; 1553 ts2->ts_ltick = ts->ts_ltick; 1554 ts2->ts_ftick = ts->ts_ftick; 1555 child->td_user_pri = td->td_user_pri; 1556 child->td_base_user_pri = td->td_base_user_pri; 1557 /* 1558 * And update interactivity score. 1559 */ 1560 ts2->skg_slptime = ts->skg_slptime; 1561 ts2->skg_runtime = ts->skg_runtime; 1562 ts2->ts_slice = 1; /* Attempt to quickly learn interactivity. */ 1563 } 1564 1565 void 1566 sched_class(struct thread *td, int class) 1567 { 1568 1569 mtx_assert(&sched_lock, MA_OWNED); 1570 if (td->td_pri_class == class) 1571 return; 1572 1573 #ifdef SMP 1574 /* 1575 * On SMP if we're on the RUNQ we must adjust the transferable 1576 * count because could be changing to or from an interrupt 1577 * class. 1578 */ 1579 if (TD_ON_RUNQ(td)) { 1580 struct tdq *tdq; 1581 1582 tdq = TDQ_CPU(td->td_sched->ts_cpu); 1583 if (THREAD_CAN_MIGRATE(td)) { 1584 tdq->tdq_transferable--; 1585 tdq->tdq_group->tdg_transferable--; 1586 } 1587 td->td_pri_class = class; 1588 if (THREAD_CAN_MIGRATE(td)) { 1589 tdq->tdq_transferable++; 1590 tdq->tdq_group->tdg_transferable++; 1591 } 1592 } 1593 #endif 1594 td->td_pri_class = class; 1595 } 1596 1597 /* 1598 * Return some of the child's priority and interactivity to the parent. 1599 */ 1600 void 1601 sched_exit(struct proc *p, struct thread *child) 1602 { 1603 struct thread *td; 1604 1605 CTR3(KTR_SCHED, "sched_exit: %p(%s) prio %d", 1606 child, child->td_proc->p_comm, child->td_priority); 1607 1608 td = FIRST_THREAD_IN_PROC(p); 1609 sched_exit_thread(td, child); 1610 } 1611 1612 void 1613 sched_exit_thread(struct thread *td, struct thread *child) 1614 { 1615 1616 CTR3(KTR_SCHED, "sched_exit_thread: %p(%s) prio %d", 1617 child, child->td_proc->p_comm, child->td_priority); 1618 1619 tdq_load_rem(TDQ_CPU(child->td_sched->ts_cpu), child->td_sched); 1620 #ifdef KSE 1621 /* 1622 * KSE forks and exits so often that this penalty causes short-lived 1623 * threads to always be non-interactive. This causes mozilla to 1624 * crawl under load. 1625 */ 1626 if ((td->td_pflags & TDP_SA) && td->td_proc == child->td_proc) 1627 return; 1628 #endif 1629 /* 1630 * Give the child's runtime to the parent without returning the 1631 * sleep time as a penalty to the parent. This causes shells that 1632 * launch expensive things to mark their children as expensive. 1633 */ 1634 td->td_sched->skg_runtime += child->td_sched->skg_runtime; 1635 sched_interact_update(td); 1636 sched_priority(td); 1637 } 1638 1639 void 1640 sched_userret(struct thread *td) 1641 { 1642 /* 1643 * XXX we cheat slightly on the locking here to avoid locking in 1644 * the usual case. Setting td_priority here is essentially an 1645 * incomplete workaround for not setting it properly elsewhere. 1646 * Now that some interrupt handlers are threads, not setting it 1647 * properly elsewhere can clobber it in the window between setting 1648 * it here and returning to user mode, so don't waste time setting 1649 * it perfectly here. 1650 */ 1651 KASSERT((td->td_flags & TDF_BORROWING) == 0, 1652 ("thread with borrowed priority returning to userland")); 1653 if (td->td_priority != td->td_user_pri) { 1654 mtx_lock_spin(&sched_lock); 1655 td->td_priority = td->td_user_pri; 1656 td->td_base_pri = td->td_user_pri; 1657 mtx_unlock_spin(&sched_lock); 1658 } 1659 } 1660 1661 void 1662 sched_clock(struct thread *td) 1663 { 1664 struct tdq *tdq; 1665 struct td_sched *ts; 1666 1667 mtx_assert(&sched_lock, MA_OWNED); 1668 #ifdef SMP 1669 sched_smp_tick(td); 1670 #endif 1671 tdq = TDQ_SELF(); 1672 /* 1673 * Advance the insert index once for each tick to ensure that all 1674 * threads get a chance to run. 1675 */ 1676 if (tdq->tdq_idx == tdq->tdq_ridx) { 1677 tdq->tdq_idx = (tdq->tdq_idx + 1) % RQ_NQS; 1678 if (TAILQ_EMPTY(&tdq->tdq_timeshare.rq_queues[tdq->tdq_ridx])) 1679 tdq->tdq_ridx = tdq->tdq_idx; 1680 } 1681 ts = td->td_sched; 1682 /* 1683 * We only do slicing code for TIMESHARE threads. 1684 */ 1685 if (td->td_pri_class != PRI_TIMESHARE) 1686 return; 1687 /* 1688 * We used a tick; charge it to the thread so that we can compute our 1689 * interactivity. 1690 */ 1691 td->td_sched->skg_runtime += tickincr; 1692 sched_interact_update(td); 1693 /* 1694 * We used up one time slice. 1695 */ 1696 if (--ts->ts_slice > 0) 1697 return; 1698 /* 1699 * We're out of time, recompute priorities and requeue. 1700 */ 1701 sched_priority(td); 1702 td->td_flags |= TDF_NEEDRESCHED; 1703 } 1704 1705 int 1706 sched_runnable(void) 1707 { 1708 struct tdq *tdq; 1709 int load; 1710 1711 load = 1; 1712 1713 tdq = TDQ_SELF(); 1714 #ifdef SMP 1715 if (tdq_busy) 1716 goto out; 1717 #endif 1718 if ((curthread->td_flags & TDF_IDLETD) != 0) { 1719 if (tdq->tdq_load > 0) 1720 goto out; 1721 } else 1722 if (tdq->tdq_load - 1 > 0) 1723 goto out; 1724 load = 0; 1725 out: 1726 return (load); 1727 } 1728 1729 struct thread * 1730 sched_choose(void) 1731 { 1732 struct tdq *tdq; 1733 struct td_sched *ts; 1734 1735 mtx_assert(&sched_lock, MA_OWNED); 1736 tdq = TDQ_SELF(); 1737 #ifdef SMP 1738 restart: 1739 #endif 1740 ts = tdq_choose(tdq); 1741 if (ts) { 1742 #ifdef SMP 1743 if (ts->ts_thread->td_priority > PRI_MIN_IDLE) 1744 if (tdq_idled(tdq) == 0) 1745 goto restart; 1746 #endif 1747 tdq_runq_rem(tdq, ts); 1748 return (ts->ts_thread); 1749 } 1750 #ifdef SMP 1751 if (tdq_idled(tdq) == 0) 1752 goto restart; 1753 #endif 1754 return (PCPU_GET(idlethread)); 1755 } 1756 1757 static int 1758 sched_preempt(struct thread *td) 1759 { 1760 struct thread *ctd; 1761 int cpri; 1762 int pri; 1763 1764 ctd = curthread; 1765 pri = td->td_priority; 1766 cpri = ctd->td_priority; 1767 if (panicstr != NULL || pri >= cpri || cold || TD_IS_INHIBITED(ctd)) 1768 return (0); 1769 /* 1770 * Always preempt IDLE threads. Otherwise only if the preempting 1771 * thread is an ithread. 1772 */ 1773 if (pri > PRI_MAX_ITHD && cpri < PRI_MIN_IDLE) 1774 return (0); 1775 if (ctd->td_critnest > 1) { 1776 CTR1(KTR_PROC, "sched_preempt: in critical section %d", 1777 ctd->td_critnest); 1778 ctd->td_owepreempt = 1; 1779 return (0); 1780 } 1781 /* 1782 * Thread is runnable but not yet put on system run queue. 1783 */ 1784 MPASS(TD_ON_RUNQ(td)); 1785 TD_SET_RUNNING(td); 1786 CTR3(KTR_PROC, "preempting to thread %p (pid %d, %s)\n", td, 1787 td->td_proc->p_pid, td->td_proc->p_comm); 1788 mi_switch(SW_INVOL|SW_PREEMPT, td); 1789 return (1); 1790 } 1791 1792 void 1793 sched_add(struct thread *td, int flags) 1794 { 1795 struct tdq *tdq; 1796 struct td_sched *ts; 1797 int preemptive; 1798 int class; 1799 #ifdef SMP 1800 int cpuid; 1801 int cpumask; 1802 #endif 1803 ts = td->td_sched; 1804 1805 mtx_assert(&sched_lock, MA_OWNED); 1806 CTR5(KTR_SCHED, "sched_add: %p(%s) prio %d by %p(%s)", 1807 td, td->td_proc->p_comm, td->td_priority, curthread, 1808 curthread->td_proc->p_comm); 1809 KASSERT((td->td_inhibitors == 0), 1810 ("sched_add: trying to run inhibited thread")); 1811 KASSERT((TD_CAN_RUN(td) || TD_IS_RUNNING(td)), 1812 ("sched_add: bad thread state")); 1813 KASSERT(td->td_proc->p_sflag & PS_INMEM, 1814 ("sched_add: process swapped out")); 1815 KASSERT(ts->ts_runq == NULL, 1816 ("sched_add: thread %p is still assigned to a run queue", td)); 1817 TD_SET_RUNQ(td); 1818 tdq = TDQ_SELF(); 1819 class = PRI_BASE(td->td_pri_class); 1820 preemptive = !(flags & SRQ_YIELDING); 1821 /* 1822 * Recalculate the priority before we select the target cpu or 1823 * run-queue. 1824 */ 1825 if (class == PRI_TIMESHARE) 1826 sched_priority(td); 1827 if (ts->ts_slice == 0) 1828 ts->ts_slice = sched_slice; 1829 #ifdef SMP 1830 cpuid = PCPU_GET(cpuid); 1831 /* 1832 * Pick the destination cpu and if it isn't ours transfer to the 1833 * target cpu. 1834 */ 1835 if (THREAD_CAN_MIGRATE(td)) { 1836 if (td->td_priority <= PRI_MAX_ITHD) { 1837 CTR2(KTR_SCHED, "ithd %d < %d", td->td_priority, PRI_MAX_ITHD); 1838 ts->ts_cpu = cpuid; 1839 } 1840 if (pick_pri) 1841 ts->ts_cpu = tdq_pickpri(tdq, ts, flags); 1842 else 1843 ts->ts_cpu = tdq_pickidle(tdq, ts); 1844 } else 1845 CTR1(KTR_SCHED, "pinned %d", td->td_pinned); 1846 if (ts->ts_cpu != cpuid) 1847 preemptive = 0; 1848 tdq = TDQ_CPU(ts->ts_cpu); 1849 cpumask = 1 << ts->ts_cpu; 1850 /* 1851 * If we had been idle, clear our bit in the group and potentially 1852 * the global bitmap. 1853 */ 1854 if ((class != PRI_IDLE && class != PRI_ITHD) && 1855 (tdq->tdq_group->tdg_idlemask & cpumask) != 0) { 1856 /* 1857 * Check to see if our group is unidling, and if so, remove it 1858 * from the global idle mask. 1859 */ 1860 if (tdq->tdq_group->tdg_idlemask == 1861 tdq->tdq_group->tdg_cpumask) 1862 atomic_clear_int(&tdq_idle, tdq->tdq_group->tdg_mask); 1863 /* 1864 * Now remove ourselves from the group specific idle mask. 1865 */ 1866 tdq->tdq_group->tdg_idlemask &= ~cpumask; 1867 } 1868 #endif 1869 /* 1870 * Pick the run queue based on priority. 1871 */ 1872 if (td->td_priority <= PRI_MAX_REALTIME) 1873 ts->ts_runq = &tdq->tdq_realtime; 1874 else if (td->td_priority <= PRI_MAX_TIMESHARE) 1875 ts->ts_runq = &tdq->tdq_timeshare; 1876 else 1877 ts->ts_runq = &tdq->tdq_idle; 1878 if (preemptive && sched_preempt(td)) 1879 return; 1880 tdq_runq_add(tdq, ts, flags); 1881 tdq_load_add(tdq, ts); 1882 #ifdef SMP 1883 if (ts->ts_cpu != cpuid) { 1884 tdq_notify(ts); 1885 return; 1886 } 1887 #endif 1888 if (td->td_priority < curthread->td_priority) 1889 curthread->td_flags |= TDF_NEEDRESCHED; 1890 } 1891 1892 void 1893 sched_rem(struct thread *td) 1894 { 1895 struct tdq *tdq; 1896 struct td_sched *ts; 1897 1898 CTR5(KTR_SCHED, "sched_rem: %p(%s) prio %d by %p(%s)", 1899 td, td->td_proc->p_comm, td->td_priority, curthread, 1900 curthread->td_proc->p_comm); 1901 mtx_assert(&sched_lock, MA_OWNED); 1902 ts = td->td_sched; 1903 KASSERT(TD_ON_RUNQ(td), 1904 ("sched_rem: thread not on run queue")); 1905 1906 tdq = TDQ_CPU(ts->ts_cpu); 1907 tdq_runq_rem(tdq, ts); 1908 tdq_load_rem(tdq, ts); 1909 TD_SET_CAN_RUN(td); 1910 } 1911 1912 fixpt_t 1913 sched_pctcpu(struct thread *td) 1914 { 1915 fixpt_t pctcpu; 1916 struct td_sched *ts; 1917 1918 pctcpu = 0; 1919 ts = td->td_sched; 1920 if (ts == NULL) 1921 return (0); 1922 1923 mtx_lock_spin(&sched_lock); 1924 if (ts->ts_ticks) { 1925 int rtick; 1926 1927 sched_pctcpu_update(ts); 1928 /* How many rtick per second ? */ 1929 rtick = min(SCHED_TICK_HZ(ts) / SCHED_TICK_SECS, hz); 1930 pctcpu = (FSCALE * ((FSCALE * rtick)/hz)) >> FSHIFT; 1931 } 1932 td->td_proc->p_swtime = ts->ts_ltick - ts->ts_ftick; 1933 mtx_unlock_spin(&sched_lock); 1934 1935 return (pctcpu); 1936 } 1937 1938 void 1939 sched_bind(struct thread *td, int cpu) 1940 { 1941 struct td_sched *ts; 1942 1943 mtx_assert(&sched_lock, MA_OWNED); 1944 ts = td->td_sched; 1945 if (ts->ts_flags & TSF_BOUND) 1946 sched_unbind(td); 1947 ts->ts_flags |= TSF_BOUND; 1948 #ifdef SMP 1949 sched_pin(); 1950 if (PCPU_GET(cpuid) == cpu) 1951 return; 1952 ts->ts_cpu = cpu; 1953 /* When we return from mi_switch we'll be on the correct cpu. */ 1954 mi_switch(SW_VOL, NULL); 1955 #endif 1956 } 1957 1958 void 1959 sched_unbind(struct thread *td) 1960 { 1961 struct td_sched *ts; 1962 1963 mtx_assert(&sched_lock, MA_OWNED); 1964 ts = td->td_sched; 1965 if ((ts->ts_flags & TSF_BOUND) == 0) 1966 return; 1967 ts->ts_flags &= ~TSF_BOUND; 1968 #ifdef SMP 1969 sched_unpin(); 1970 #endif 1971 } 1972 1973 int 1974 sched_is_bound(struct thread *td) 1975 { 1976 mtx_assert(&sched_lock, MA_OWNED); 1977 return (td->td_sched->ts_flags & TSF_BOUND); 1978 } 1979 1980 void 1981 sched_relinquish(struct thread *td) 1982 { 1983 mtx_lock_spin(&sched_lock); 1984 if (td->td_pri_class == PRI_TIMESHARE) 1985 sched_prio(td, PRI_MAX_TIMESHARE); 1986 mi_switch(SW_VOL, NULL); 1987 mtx_unlock_spin(&sched_lock); 1988 } 1989 1990 int 1991 sched_load(void) 1992 { 1993 #ifdef SMP 1994 int total; 1995 int i; 1996 1997 total = 0; 1998 for (i = 0; i <= tdg_maxid; i++) 1999 total += TDQ_GROUP(i)->tdg_load; 2000 return (total); 2001 #else 2002 return (TDQ_SELF()->tdq_sysload); 2003 #endif 2004 } 2005 2006 int 2007 sched_sizeof_proc(void) 2008 { 2009 return (sizeof(struct proc)); 2010 } 2011 2012 int 2013 sched_sizeof_thread(void) 2014 { 2015 return (sizeof(struct thread) + sizeof(struct td_sched)); 2016 } 2017 2018 void 2019 sched_tick(void) 2020 { 2021 struct td_sched *ts; 2022 2023 ts = curthread->td_sched; 2024 /* Adjust ticks for pctcpu */ 2025 ts->ts_ticks += 1 << SCHED_TICK_SHIFT; 2026 ts->ts_ltick = ticks; 2027 /* 2028 * Update if we've exceeded our desired tick threshhold by over one 2029 * second. 2030 */ 2031 if (ts->ts_ftick + SCHED_TICK_MAX < ts->ts_ltick) 2032 sched_pctcpu_update(ts); 2033 } 2034 2035 /* 2036 * The actual idle process. 2037 */ 2038 void 2039 sched_idletd(void *dummy) 2040 { 2041 struct proc *p; 2042 struct thread *td; 2043 2044 td = curthread; 2045 p = td->td_proc; 2046 mtx_assert(&Giant, MA_NOTOWNED); 2047 /* ULE Relies on preemption for idle interruption. */ 2048 for (;;) 2049 cpu_idle(); 2050 } 2051 2052 static SYSCTL_NODE(_kern, OID_AUTO, sched, CTLFLAG_RW, 0, "Scheduler"); 2053 SYSCTL_STRING(_kern_sched, OID_AUTO, name, CTLFLAG_RD, "ule", 0, 2054 "Scheduler name"); 2055 SYSCTL_INT(_kern_sched, OID_AUTO, slice, CTLFLAG_RW, &sched_slice, 0, ""); 2056 SYSCTL_INT(_kern_sched, OID_AUTO, interact, CTLFLAG_RW, &sched_interact, 0, ""); 2057 SYSCTL_INT(_kern_sched, OID_AUTO, tickincr, CTLFLAG_RD, &tickincr, 0, ""); 2058 SYSCTL_INT(_kern_sched, OID_AUTO, realstathz, CTLFLAG_RD, &realstathz, 0, ""); 2059 #ifdef SMP 2060 SYSCTL_INT(_kern_sched, OID_AUTO, pick_pri, CTLFLAG_RW, &pick_pri, 0, ""); 2061 SYSCTL_INT(_kern_sched, OID_AUTO, pick_pri_affinity, CTLFLAG_RW, 2062 &affinity, 0, ""); 2063 SYSCTL_INT(_kern_sched, OID_AUTO, pick_pri_tryself, CTLFLAG_RW, 2064 &tryself, 0, ""); 2065 SYSCTL_INT(_kern_sched, OID_AUTO, pick_pri_tryselfidle, CTLFLAG_RW, 2066 &tryselfidle, 0, ""); 2067 SYSCTL_INT(_kern_sched, OID_AUTO, balance, CTLFLAG_RW, &rebalance, 0, ""); 2068 SYSCTL_INT(_kern_sched, OID_AUTO, ipi_preempt, CTLFLAG_RW, &ipi_preempt, 0, ""); 2069 SYSCTL_INT(_kern_sched, OID_AUTO, ipi_ast, CTLFLAG_RW, &ipi_ast, 0, ""); 2070 SYSCTL_INT(_kern_sched, OID_AUTO, ipi_thresh, CTLFLAG_RW, &ipi_thresh, 0, ""); 2071 SYSCTL_INT(_kern_sched, OID_AUTO, steal_htt, CTLFLAG_RW, &steal_htt, 0, ""); 2072 SYSCTL_INT(_kern_sched, OID_AUTO, steal_busy, CTLFLAG_RW, &steal_busy, 0, ""); 2073 SYSCTL_INT(_kern_sched, OID_AUTO, busy_thresh, CTLFLAG_RW, &busy_thresh, 0, ""); 2074 #endif 2075 2076 /* ps compat */ 2077 static fixpt_t ccpu = 0.95122942450071400909 * FSCALE; /* exp(-1/20) */ 2078 SYSCTL_INT(_kern, OID_AUTO, ccpu, CTLFLAG_RD, &ccpu, 0, ""); 2079 2080 2081 #define KERN_SWITCH_INCLUDE 1 2082 #include "kern/kern_switch.c" 2083