1 /*- 2 * Copyright (c) 1982, 1986, 1990, 1991, 1993 3 * The Regents of the University of California. All rights reserved. 4 * (c) UNIX System Laboratories, Inc. 5 * All or some portions of this file are derived from material licensed 6 * to the University of California by American Telephone and Telegraph 7 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 8 * the permission of UNIX System Laboratories, Inc. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 4. Neither the name of the University nor the names of its contributors 19 * may be used to endorse or promote products derived from this software 20 * without specific prior written permission. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 32 * SUCH DAMAGE. 33 */ 34 35 #include <sys/cdefs.h> 36 __FBSDID("$FreeBSD$"); 37 38 #include "opt_hwpmc_hooks.h" 39 #include "opt_sched.h" 40 #include "opt_kdtrace.h" 41 42 #include <sys/param.h> 43 #include <sys/systm.h> 44 #include <sys/cpuset.h> 45 #include <sys/kernel.h> 46 #include <sys/ktr.h> 47 #include <sys/lock.h> 48 #include <sys/kthread.h> 49 #include <sys/mutex.h> 50 #include <sys/proc.h> 51 #include <sys/resourcevar.h> 52 #include <sys/sched.h> 53 #include <sys/smp.h> 54 #include <sys/sysctl.h> 55 #include <sys/sx.h> 56 #include <sys/turnstile.h> 57 #include <sys/umtx.h> 58 #include <machine/pcb.h> 59 #include <machine/smp.h> 60 61 #ifdef HWPMC_HOOKS 62 #include <sys/pmckern.h> 63 #endif 64 65 #ifdef KDTRACE_HOOKS 66 #include <sys/dtrace_bsd.h> 67 int dtrace_vtime_active; 68 dtrace_vtime_switch_func_t dtrace_vtime_switch_func; 69 #endif 70 71 /* 72 * INVERSE_ESTCPU_WEIGHT is only suitable for statclock() frequencies in 73 * the range 100-256 Hz (approximately). 74 */ 75 #define ESTCPULIM(e) \ 76 min((e), INVERSE_ESTCPU_WEIGHT * (NICE_WEIGHT * (PRIO_MAX - PRIO_MIN) - \ 77 RQ_PPQ) + INVERSE_ESTCPU_WEIGHT - 1) 78 #ifdef SMP 79 #define INVERSE_ESTCPU_WEIGHT (8 * smp_cpus) 80 #else 81 #define INVERSE_ESTCPU_WEIGHT 8 /* 1 / (priorities per estcpu level). */ 82 #endif 83 #define NICE_WEIGHT 1 /* Priorities per nice level. */ 84 85 #define TS_NAME_LEN (MAXCOMLEN + sizeof(" td ") + sizeof(__XSTRING(UINT_MAX))) 86 87 /* 88 * The schedulable entity that runs a context. 89 * This is an extension to the thread structure and is tailored to 90 * the requirements of this scheduler 91 */ 92 struct td_sched { 93 fixpt_t ts_pctcpu; /* (j) %cpu during p_swtime. */ 94 int ts_cpticks; /* (j) Ticks of cpu time. */ 95 int ts_slptime; /* (j) Seconds !RUNNING. */ 96 int ts_flags; 97 struct runq *ts_runq; /* runq the thread is currently on */ 98 #ifdef KTR 99 char ts_name[TS_NAME_LEN]; 100 #endif 101 }; 102 103 /* flags kept in td_flags */ 104 #define TDF_DIDRUN TDF_SCHED0 /* thread actually ran. */ 105 #define TDF_BOUND TDF_SCHED1 /* Bound to one CPU. */ 106 107 /* flags kept in ts_flags */ 108 #define TSF_AFFINITY 0x0001 /* Has a non-"full" CPU set. */ 109 110 #define SKE_RUNQ_PCPU(ts) \ 111 ((ts)->ts_runq != 0 && (ts)->ts_runq != &runq) 112 113 #define THREAD_CAN_SCHED(td, cpu) \ 114 CPU_ISSET((cpu), &(td)->td_cpuset->cs_mask) 115 116 static struct td_sched td_sched0; 117 struct mtx sched_lock; 118 119 static int sched_tdcnt; /* Total runnable threads in the system. */ 120 static int sched_quantum; /* Roundrobin scheduling quantum in ticks. */ 121 #define SCHED_QUANTUM (hz / 10) /* Default sched quantum */ 122 123 static void setup_runqs(void); 124 static void schedcpu(void); 125 static void schedcpu_thread(void); 126 static void sched_priority(struct thread *td, u_char prio); 127 static void sched_setup(void *dummy); 128 static void maybe_resched(struct thread *td); 129 static void updatepri(struct thread *td); 130 static void resetpriority(struct thread *td); 131 static void resetpriority_thread(struct thread *td); 132 #ifdef SMP 133 static int sched_pickcpu(struct thread *td); 134 static int forward_wakeup(int cpunum); 135 static void kick_other_cpu(int pri, int cpuid); 136 #endif 137 138 static struct kproc_desc sched_kp = { 139 "schedcpu", 140 schedcpu_thread, 141 NULL 142 }; 143 SYSINIT(schedcpu, SI_SUB_RUN_SCHEDULER, SI_ORDER_FIRST, kproc_start, 144 &sched_kp); 145 SYSINIT(sched_setup, SI_SUB_RUN_QUEUE, SI_ORDER_FIRST, sched_setup, NULL); 146 147 /* 148 * Global run queue. 149 */ 150 static struct runq runq; 151 152 #ifdef SMP 153 /* 154 * Per-CPU run queues 155 */ 156 static struct runq runq_pcpu[MAXCPU]; 157 long runq_length[MAXCPU]; 158 #endif 159 160 struct pcpuidlestat { 161 u_int idlecalls; 162 u_int oldidlecalls; 163 }; 164 static DPCPU_DEFINE(struct pcpuidlestat, idlestat); 165 166 static void 167 setup_runqs(void) 168 { 169 #ifdef SMP 170 int i; 171 172 for (i = 0; i < MAXCPU; ++i) 173 runq_init(&runq_pcpu[i]); 174 #endif 175 176 runq_init(&runq); 177 } 178 179 static int 180 sysctl_kern_quantum(SYSCTL_HANDLER_ARGS) 181 { 182 int error, new_val; 183 184 new_val = sched_quantum * tick; 185 error = sysctl_handle_int(oidp, &new_val, 0, req); 186 if (error != 0 || req->newptr == NULL) 187 return (error); 188 if (new_val < tick) 189 return (EINVAL); 190 sched_quantum = new_val / tick; 191 hogticks = 2 * sched_quantum; 192 return (0); 193 } 194 195 SYSCTL_NODE(_kern, OID_AUTO, sched, CTLFLAG_RD, 0, "Scheduler"); 196 197 SYSCTL_STRING(_kern_sched, OID_AUTO, name, CTLFLAG_RD, "4BSD", 0, 198 "Scheduler name"); 199 200 SYSCTL_PROC(_kern_sched, OID_AUTO, quantum, CTLTYPE_INT | CTLFLAG_RW, 201 0, sizeof sched_quantum, sysctl_kern_quantum, "I", 202 "Roundrobin scheduling quantum in microseconds"); 203 204 #ifdef SMP 205 /* Enable forwarding of wakeups to all other cpus */ 206 SYSCTL_NODE(_kern_sched, OID_AUTO, ipiwakeup, CTLFLAG_RD, NULL, "Kernel SMP"); 207 208 static int runq_fuzz = 1; 209 SYSCTL_INT(_kern_sched, OID_AUTO, runq_fuzz, CTLFLAG_RW, &runq_fuzz, 0, ""); 210 211 static int forward_wakeup_enabled = 1; 212 SYSCTL_INT(_kern_sched_ipiwakeup, OID_AUTO, enabled, CTLFLAG_RW, 213 &forward_wakeup_enabled, 0, 214 "Forwarding of wakeup to idle CPUs"); 215 216 static int forward_wakeups_requested = 0; 217 SYSCTL_INT(_kern_sched_ipiwakeup, OID_AUTO, requested, CTLFLAG_RD, 218 &forward_wakeups_requested, 0, 219 "Requests for Forwarding of wakeup to idle CPUs"); 220 221 static int forward_wakeups_delivered = 0; 222 SYSCTL_INT(_kern_sched_ipiwakeup, OID_AUTO, delivered, CTLFLAG_RD, 223 &forward_wakeups_delivered, 0, 224 "Completed Forwarding of wakeup to idle CPUs"); 225 226 static int forward_wakeup_use_mask = 1; 227 SYSCTL_INT(_kern_sched_ipiwakeup, OID_AUTO, usemask, CTLFLAG_RW, 228 &forward_wakeup_use_mask, 0, 229 "Use the mask of idle cpus"); 230 231 static int forward_wakeup_use_loop = 0; 232 SYSCTL_INT(_kern_sched_ipiwakeup, OID_AUTO, useloop, CTLFLAG_RW, 233 &forward_wakeup_use_loop, 0, 234 "Use a loop to find idle cpus"); 235 236 static int forward_wakeup_use_single = 0; 237 SYSCTL_INT(_kern_sched_ipiwakeup, OID_AUTO, onecpu, CTLFLAG_RW, 238 &forward_wakeup_use_single, 0, 239 "Only signal one idle cpu"); 240 241 static int forward_wakeup_use_htt = 0; 242 SYSCTL_INT(_kern_sched_ipiwakeup, OID_AUTO, htt2, CTLFLAG_RW, 243 &forward_wakeup_use_htt, 0, 244 "account for htt"); 245 246 #endif 247 #if 0 248 static int sched_followon = 0; 249 SYSCTL_INT(_kern_sched, OID_AUTO, followon, CTLFLAG_RW, 250 &sched_followon, 0, 251 "allow threads to share a quantum"); 252 #endif 253 254 static __inline void 255 sched_load_add(void) 256 { 257 258 sched_tdcnt++; 259 KTR_COUNTER0(KTR_SCHED, "load", "global load", sched_tdcnt); 260 } 261 262 static __inline void 263 sched_load_rem(void) 264 { 265 266 sched_tdcnt--; 267 KTR_COUNTER0(KTR_SCHED, "load", "global load", sched_tdcnt); 268 } 269 /* 270 * Arrange to reschedule if necessary, taking the priorities and 271 * schedulers into account. 272 */ 273 static void 274 maybe_resched(struct thread *td) 275 { 276 277 THREAD_LOCK_ASSERT(td, MA_OWNED); 278 if (td->td_priority < curthread->td_priority) 279 curthread->td_flags |= TDF_NEEDRESCHED; 280 } 281 282 /* 283 * This function is called when a thread is about to be put on run queue 284 * because it has been made runnable or its priority has been adjusted. It 285 * determines if the new thread should be immediately preempted to. If so, 286 * it switches to it and eventually returns true. If not, it returns false 287 * so that the caller may place the thread on an appropriate run queue. 288 */ 289 int 290 maybe_preempt(struct thread *td) 291 { 292 #ifdef PREEMPTION 293 struct thread *ctd; 294 int cpri, pri; 295 296 /* 297 * The new thread should not preempt the current thread if any of the 298 * following conditions are true: 299 * 300 * - The kernel is in the throes of crashing (panicstr). 301 * - The current thread has a higher (numerically lower) or 302 * equivalent priority. Note that this prevents curthread from 303 * trying to preempt to itself. 304 * - It is too early in the boot for context switches (cold is set). 305 * - The current thread has an inhibitor set or is in the process of 306 * exiting. In this case, the current thread is about to switch 307 * out anyways, so there's no point in preempting. If we did, 308 * the current thread would not be properly resumed as well, so 309 * just avoid that whole landmine. 310 * - If the new thread's priority is not a realtime priority and 311 * the current thread's priority is not an idle priority and 312 * FULL_PREEMPTION is disabled. 313 * 314 * If all of these conditions are false, but the current thread is in 315 * a nested critical section, then we have to defer the preemption 316 * until we exit the critical section. Otherwise, switch immediately 317 * to the new thread. 318 */ 319 ctd = curthread; 320 THREAD_LOCK_ASSERT(td, MA_OWNED); 321 KASSERT((td->td_inhibitors == 0), 322 ("maybe_preempt: trying to run inhibited thread")); 323 pri = td->td_priority; 324 cpri = ctd->td_priority; 325 if (panicstr != NULL || pri >= cpri || cold /* || dumping */ || 326 TD_IS_INHIBITED(ctd)) 327 return (0); 328 #ifndef FULL_PREEMPTION 329 if (pri > PRI_MAX_ITHD && cpri < PRI_MIN_IDLE) 330 return (0); 331 #endif 332 333 if (ctd->td_critnest > 1) { 334 CTR1(KTR_PROC, "maybe_preempt: in critical section %d", 335 ctd->td_critnest); 336 ctd->td_owepreempt = 1; 337 return (0); 338 } 339 /* 340 * Thread is runnable but not yet put on system run queue. 341 */ 342 MPASS(ctd->td_lock == td->td_lock); 343 MPASS(TD_ON_RUNQ(td)); 344 TD_SET_RUNNING(td); 345 CTR3(KTR_PROC, "preempting to thread %p (pid %d, %s)\n", td, 346 td->td_proc->p_pid, td->td_name); 347 mi_switch(SW_INVOL | SW_PREEMPT | SWT_PREEMPT, td); 348 /* 349 * td's lock pointer may have changed. We have to return with it 350 * locked. 351 */ 352 spinlock_enter(); 353 thread_unlock(ctd); 354 thread_lock(td); 355 spinlock_exit(); 356 return (1); 357 #else 358 return (0); 359 #endif 360 } 361 362 /* 363 * Constants for digital decay and forget: 364 * 90% of (td_estcpu) usage in 5 * loadav time 365 * 95% of (ts_pctcpu) usage in 60 seconds (load insensitive) 366 * Note that, as ps(1) mentions, this can let percentages 367 * total over 100% (I've seen 137.9% for 3 processes). 368 * 369 * Note that schedclock() updates td_estcpu and p_cpticks asynchronously. 370 * 371 * We wish to decay away 90% of td_estcpu in (5 * loadavg) seconds. 372 * That is, the system wants to compute a value of decay such 373 * that the following for loop: 374 * for (i = 0; i < (5 * loadavg); i++) 375 * td_estcpu *= decay; 376 * will compute 377 * td_estcpu *= 0.1; 378 * for all values of loadavg: 379 * 380 * Mathematically this loop can be expressed by saying: 381 * decay ** (5 * loadavg) ~= .1 382 * 383 * The system computes decay as: 384 * decay = (2 * loadavg) / (2 * loadavg + 1) 385 * 386 * We wish to prove that the system's computation of decay 387 * will always fulfill the equation: 388 * decay ** (5 * loadavg) ~= .1 389 * 390 * If we compute b as: 391 * b = 2 * loadavg 392 * then 393 * decay = b / (b + 1) 394 * 395 * We now need to prove two things: 396 * 1) Given factor ** (5 * loadavg) ~= .1, prove factor == b/(b+1) 397 * 2) Given b/(b+1) ** power ~= .1, prove power == (5 * loadavg) 398 * 399 * Facts: 400 * For x close to zero, exp(x) =~ 1 + x, since 401 * exp(x) = 0! + x**1/1! + x**2/2! + ... . 402 * therefore exp(-1/b) =~ 1 - (1/b) = (b-1)/b. 403 * For x close to zero, ln(1+x) =~ x, since 404 * ln(1+x) = x - x**2/2 + x**3/3 - ... -1 < x < 1 405 * therefore ln(b/(b+1)) = ln(1 - 1/(b+1)) =~ -1/(b+1). 406 * ln(.1) =~ -2.30 407 * 408 * Proof of (1): 409 * Solve (factor)**(power) =~ .1 given power (5*loadav): 410 * solving for factor, 411 * ln(factor) =~ (-2.30/5*loadav), or 412 * factor =~ exp(-1/((5/2.30)*loadav)) =~ exp(-1/(2*loadav)) = 413 * exp(-1/b) =~ (b-1)/b =~ b/(b+1). QED 414 * 415 * Proof of (2): 416 * Solve (factor)**(power) =~ .1 given factor == (b/(b+1)): 417 * solving for power, 418 * power*ln(b/(b+1)) =~ -2.30, or 419 * power =~ 2.3 * (b + 1) = 4.6*loadav + 2.3 =~ 5*loadav. QED 420 * 421 * Actual power values for the implemented algorithm are as follows: 422 * loadav: 1 2 3 4 423 * power: 5.68 10.32 14.94 19.55 424 */ 425 426 /* calculations for digital decay to forget 90% of usage in 5*loadav sec */ 427 #define loadfactor(loadav) (2 * (loadav)) 428 #define decay_cpu(loadfac, cpu) (((loadfac) * (cpu)) / ((loadfac) + FSCALE)) 429 430 /* decay 95% of `ts_pctcpu' in 60 seconds; see CCPU_SHIFT before changing */ 431 static fixpt_t ccpu = 0.95122942450071400909 * FSCALE; /* exp(-1/20) */ 432 SYSCTL_UINT(_kern, OID_AUTO, ccpu, CTLFLAG_RD, &ccpu, 0, ""); 433 434 /* 435 * If `ccpu' is not equal to `exp(-1/20)' and you still want to use the 436 * faster/more-accurate formula, you'll have to estimate CCPU_SHIFT below 437 * and possibly adjust FSHIFT in "param.h" so that (FSHIFT >= CCPU_SHIFT). 438 * 439 * To estimate CCPU_SHIFT for exp(-1/20), the following formula was used: 440 * 1 - exp(-1/20) ~= 0.0487 ~= 0.0488 == 1 (fixed pt, *11* bits). 441 * 442 * If you don't want to bother with the faster/more-accurate formula, you 443 * can set CCPU_SHIFT to (FSHIFT + 1) which will use a slower/less-accurate 444 * (more general) method of calculating the %age of CPU used by a process. 445 */ 446 #define CCPU_SHIFT 11 447 448 /* 449 * Recompute process priorities, every hz ticks. 450 * MP-safe, called without the Giant mutex. 451 */ 452 /* ARGSUSED */ 453 static void 454 schedcpu(void) 455 { 456 register fixpt_t loadfac = loadfactor(averunnable.ldavg[0]); 457 struct thread *td; 458 struct proc *p; 459 struct td_sched *ts; 460 int awake, realstathz; 461 462 realstathz = stathz ? stathz : hz; 463 sx_slock(&allproc_lock); 464 FOREACH_PROC_IN_SYSTEM(p) { 465 PROC_LOCK(p); 466 if (p->p_state == PRS_NEW) { 467 PROC_UNLOCK(p); 468 continue; 469 } 470 FOREACH_THREAD_IN_PROC(p, td) { 471 awake = 0; 472 thread_lock(td); 473 ts = td->td_sched; 474 /* 475 * Increment sleep time (if sleeping). We 476 * ignore overflow, as above. 477 */ 478 /* 479 * The td_sched slptimes are not touched in wakeup 480 * because the thread may not HAVE everything in 481 * memory? XXX I think this is out of date. 482 */ 483 if (TD_ON_RUNQ(td)) { 484 awake = 1; 485 td->td_flags &= ~TDF_DIDRUN; 486 } else if (TD_IS_RUNNING(td)) { 487 awake = 1; 488 /* Do not clear TDF_DIDRUN */ 489 } else if (td->td_flags & TDF_DIDRUN) { 490 awake = 1; 491 td->td_flags &= ~TDF_DIDRUN; 492 } 493 494 /* 495 * ts_pctcpu is only for ps and ttyinfo(). 496 */ 497 ts->ts_pctcpu = (ts->ts_pctcpu * ccpu) >> FSHIFT; 498 /* 499 * If the td_sched has been idle the entire second, 500 * stop recalculating its priority until 501 * it wakes up. 502 */ 503 if (ts->ts_cpticks != 0) { 504 #if (FSHIFT >= CCPU_SHIFT) 505 ts->ts_pctcpu += (realstathz == 100) 506 ? ((fixpt_t) ts->ts_cpticks) << 507 (FSHIFT - CCPU_SHIFT) : 508 100 * (((fixpt_t) ts->ts_cpticks) 509 << (FSHIFT - CCPU_SHIFT)) / realstathz; 510 #else 511 ts->ts_pctcpu += ((FSCALE - ccpu) * 512 (ts->ts_cpticks * 513 FSCALE / realstathz)) >> FSHIFT; 514 #endif 515 ts->ts_cpticks = 0; 516 } 517 /* 518 * If there are ANY running threads in this process, 519 * then don't count it as sleeping. 520 * XXX: this is broken. 521 */ 522 if (awake) { 523 if (ts->ts_slptime > 1) { 524 /* 525 * In an ideal world, this should not 526 * happen, because whoever woke us 527 * up from the long sleep should have 528 * unwound the slptime and reset our 529 * priority before we run at the stale 530 * priority. Should KASSERT at some 531 * point when all the cases are fixed. 532 */ 533 updatepri(td); 534 } 535 ts->ts_slptime = 0; 536 } else 537 ts->ts_slptime++; 538 if (ts->ts_slptime > 1) { 539 thread_unlock(td); 540 continue; 541 } 542 td->td_estcpu = decay_cpu(loadfac, td->td_estcpu); 543 resetpriority(td); 544 resetpriority_thread(td); 545 thread_unlock(td); 546 } 547 PROC_UNLOCK(p); 548 } 549 sx_sunlock(&allproc_lock); 550 } 551 552 /* 553 * Main loop for a kthread that executes schedcpu once a second. 554 */ 555 static void 556 schedcpu_thread(void) 557 { 558 559 for (;;) { 560 schedcpu(); 561 pause("-", hz); 562 } 563 } 564 565 /* 566 * Recalculate the priority of a process after it has slept for a while. 567 * For all load averages >= 1 and max td_estcpu of 255, sleeping for at 568 * least six times the loadfactor will decay td_estcpu to zero. 569 */ 570 static void 571 updatepri(struct thread *td) 572 { 573 struct td_sched *ts; 574 fixpt_t loadfac; 575 unsigned int newcpu; 576 577 ts = td->td_sched; 578 loadfac = loadfactor(averunnable.ldavg[0]); 579 if (ts->ts_slptime > 5 * loadfac) 580 td->td_estcpu = 0; 581 else { 582 newcpu = td->td_estcpu; 583 ts->ts_slptime--; /* was incremented in schedcpu() */ 584 while (newcpu && --ts->ts_slptime) 585 newcpu = decay_cpu(loadfac, newcpu); 586 td->td_estcpu = newcpu; 587 } 588 } 589 590 /* 591 * Compute the priority of a process when running in user mode. 592 * Arrange to reschedule if the resulting priority is better 593 * than that of the current process. 594 */ 595 static void 596 resetpriority(struct thread *td) 597 { 598 register unsigned int newpriority; 599 600 if (td->td_pri_class == PRI_TIMESHARE) { 601 newpriority = PUSER + td->td_estcpu / INVERSE_ESTCPU_WEIGHT + 602 NICE_WEIGHT * (td->td_proc->p_nice - PRIO_MIN); 603 newpriority = min(max(newpriority, PRI_MIN_TIMESHARE), 604 PRI_MAX_TIMESHARE); 605 sched_user_prio(td, newpriority); 606 } 607 } 608 609 /* 610 * Update the thread's priority when the associated process's user 611 * priority changes. 612 */ 613 static void 614 resetpriority_thread(struct thread *td) 615 { 616 617 /* Only change threads with a time sharing user priority. */ 618 if (td->td_priority < PRI_MIN_TIMESHARE || 619 td->td_priority > PRI_MAX_TIMESHARE) 620 return; 621 622 /* XXX the whole needresched thing is broken, but not silly. */ 623 maybe_resched(td); 624 625 sched_prio(td, td->td_user_pri); 626 } 627 628 /* ARGSUSED */ 629 static void 630 sched_setup(void *dummy) 631 { 632 setup_runqs(); 633 634 if (sched_quantum == 0) 635 sched_quantum = SCHED_QUANTUM; 636 hogticks = 2 * sched_quantum; 637 638 /* Account for thread0. */ 639 sched_load_add(); 640 } 641 642 /* External interfaces start here */ 643 644 /* 645 * Very early in the boot some setup of scheduler-specific 646 * parts of proc0 and of some scheduler resources needs to be done. 647 * Called from: 648 * proc0_init() 649 */ 650 void 651 schedinit(void) 652 { 653 /* 654 * Set up the scheduler specific parts of proc0. 655 */ 656 proc0.p_sched = NULL; /* XXX */ 657 thread0.td_sched = &td_sched0; 658 thread0.td_lock = &sched_lock; 659 mtx_init(&sched_lock, "sched lock", NULL, MTX_SPIN | MTX_RECURSE); 660 } 661 662 int 663 sched_runnable(void) 664 { 665 #ifdef SMP 666 return runq_check(&runq) + runq_check(&runq_pcpu[PCPU_GET(cpuid)]); 667 #else 668 return runq_check(&runq); 669 #endif 670 } 671 672 int 673 sched_rr_interval(void) 674 { 675 if (sched_quantum == 0) 676 sched_quantum = SCHED_QUANTUM; 677 return (sched_quantum); 678 } 679 680 /* 681 * We adjust the priority of the current process. The priority of 682 * a process gets worse as it accumulates CPU time. The cpu usage 683 * estimator (td_estcpu) is increased here. resetpriority() will 684 * compute a different priority each time td_estcpu increases by 685 * INVERSE_ESTCPU_WEIGHT 686 * (until MAXPRI is reached). The cpu usage estimator ramps up 687 * quite quickly when the process is running (linearly), and decays 688 * away exponentially, at a rate which is proportionally slower when 689 * the system is busy. The basic principle is that the system will 690 * 90% forget that the process used a lot of CPU time in 5 * loadav 691 * seconds. This causes the system to favor processes which haven't 692 * run much recently, and to round-robin among other processes. 693 */ 694 void 695 sched_clock(struct thread *td) 696 { 697 struct pcpuidlestat *stat; 698 struct td_sched *ts; 699 700 THREAD_LOCK_ASSERT(td, MA_OWNED); 701 ts = td->td_sched; 702 703 ts->ts_cpticks++; 704 td->td_estcpu = ESTCPULIM(td->td_estcpu + 1); 705 if ((td->td_estcpu % INVERSE_ESTCPU_WEIGHT) == 0) { 706 resetpriority(td); 707 resetpriority_thread(td); 708 } 709 710 /* 711 * Force a context switch if the current thread has used up a full 712 * quantum (default quantum is 100ms). 713 */ 714 if (!TD_IS_IDLETHREAD(td) && 715 ticks - PCPU_GET(switchticks) >= sched_quantum) 716 td->td_flags |= TDF_NEEDRESCHED; 717 718 stat = DPCPU_PTR(idlestat); 719 stat->oldidlecalls = stat->idlecalls; 720 stat->idlecalls = 0; 721 } 722 723 /* 724 * Charge child's scheduling CPU usage to parent. 725 */ 726 void 727 sched_exit(struct proc *p, struct thread *td) 728 { 729 730 KTR_STATE1(KTR_SCHED, "thread", sched_tdname(td), "proc exit", 731 "prio:td", td->td_priority); 732 733 PROC_LOCK_ASSERT(p, MA_OWNED); 734 sched_exit_thread(FIRST_THREAD_IN_PROC(p), td); 735 } 736 737 void 738 sched_exit_thread(struct thread *td, struct thread *child) 739 { 740 741 KTR_STATE1(KTR_SCHED, "thread", sched_tdname(child), "exit", 742 "prio:td", child->td_priority); 743 thread_lock(td); 744 td->td_estcpu = ESTCPULIM(td->td_estcpu + child->td_estcpu); 745 thread_unlock(td); 746 thread_lock(child); 747 if ((child->td_flags & TDF_NOLOAD) == 0) 748 sched_load_rem(); 749 thread_unlock(child); 750 } 751 752 void 753 sched_fork(struct thread *td, struct thread *childtd) 754 { 755 sched_fork_thread(td, childtd); 756 } 757 758 void 759 sched_fork_thread(struct thread *td, struct thread *childtd) 760 { 761 struct td_sched *ts; 762 763 childtd->td_estcpu = td->td_estcpu; 764 childtd->td_lock = &sched_lock; 765 childtd->td_cpuset = cpuset_ref(td->td_cpuset); 766 childtd->td_priority = childtd->td_base_pri; 767 ts = childtd->td_sched; 768 bzero(ts, sizeof(*ts)); 769 ts->ts_flags |= (td->td_sched->ts_flags & TSF_AFFINITY); 770 } 771 772 void 773 sched_nice(struct proc *p, int nice) 774 { 775 struct thread *td; 776 777 PROC_LOCK_ASSERT(p, MA_OWNED); 778 p->p_nice = nice; 779 FOREACH_THREAD_IN_PROC(p, td) { 780 thread_lock(td); 781 resetpriority(td); 782 resetpriority_thread(td); 783 thread_unlock(td); 784 } 785 } 786 787 void 788 sched_class(struct thread *td, int class) 789 { 790 THREAD_LOCK_ASSERT(td, MA_OWNED); 791 td->td_pri_class = class; 792 } 793 794 /* 795 * Adjust the priority of a thread. 796 */ 797 static void 798 sched_priority(struct thread *td, u_char prio) 799 { 800 801 802 KTR_POINT3(KTR_SCHED, "thread", sched_tdname(td), "priority change", 803 "prio:%d", td->td_priority, "new prio:%d", prio, KTR_ATTR_LINKED, 804 sched_tdname(curthread)); 805 if (td != curthread && prio > td->td_priority) { 806 KTR_POINT3(KTR_SCHED, "thread", sched_tdname(curthread), 807 "lend prio", "prio:%d", td->td_priority, "new prio:%d", 808 prio, KTR_ATTR_LINKED, sched_tdname(td)); 809 } 810 THREAD_LOCK_ASSERT(td, MA_OWNED); 811 if (td->td_priority == prio) 812 return; 813 td->td_priority = prio; 814 if (TD_ON_RUNQ(td) && td->td_rqindex != (prio / RQ_PPQ)) { 815 sched_rem(td); 816 sched_add(td, SRQ_BORING); 817 } 818 } 819 820 /* 821 * Update a thread's priority when it is lent another thread's 822 * priority. 823 */ 824 void 825 sched_lend_prio(struct thread *td, u_char prio) 826 { 827 828 td->td_flags |= TDF_BORROWING; 829 sched_priority(td, prio); 830 } 831 832 /* 833 * Restore a thread's priority when priority propagation is 834 * over. The prio argument is the minimum priority the thread 835 * needs to have to satisfy other possible priority lending 836 * requests. If the thread's regulary priority is less 837 * important than prio the thread will keep a priority boost 838 * of prio. 839 */ 840 void 841 sched_unlend_prio(struct thread *td, u_char prio) 842 { 843 u_char base_pri; 844 845 if (td->td_base_pri >= PRI_MIN_TIMESHARE && 846 td->td_base_pri <= PRI_MAX_TIMESHARE) 847 base_pri = td->td_user_pri; 848 else 849 base_pri = td->td_base_pri; 850 if (prio >= base_pri) { 851 td->td_flags &= ~TDF_BORROWING; 852 sched_prio(td, base_pri); 853 } else 854 sched_lend_prio(td, prio); 855 } 856 857 void 858 sched_prio(struct thread *td, u_char prio) 859 { 860 u_char oldprio; 861 862 /* First, update the base priority. */ 863 td->td_base_pri = prio; 864 865 /* 866 * If the thread is borrowing another thread's priority, don't ever 867 * lower the priority. 868 */ 869 if (td->td_flags & TDF_BORROWING && td->td_priority < prio) 870 return; 871 872 /* Change the real priority. */ 873 oldprio = td->td_priority; 874 sched_priority(td, prio); 875 876 /* 877 * If the thread is on a turnstile, then let the turnstile update 878 * its state. 879 */ 880 if (TD_ON_LOCK(td) && oldprio != prio) 881 turnstile_adjust(td, oldprio); 882 } 883 884 void 885 sched_user_prio(struct thread *td, u_char prio) 886 { 887 888 THREAD_LOCK_ASSERT(td, MA_OWNED); 889 td->td_base_user_pri = prio; 890 if (td->td_lend_user_pri <= prio) 891 return; 892 td->td_user_pri = prio; 893 } 894 895 void 896 sched_lend_user_prio(struct thread *td, u_char prio) 897 { 898 899 THREAD_LOCK_ASSERT(td, MA_OWNED); 900 td->td_lend_user_pri = prio; 901 td->td_user_pri = min(prio, td->td_base_user_pri); 902 if (td->td_priority > td->td_user_pri) 903 sched_prio(td, td->td_user_pri); 904 else if (td->td_priority != td->td_user_pri) 905 td->td_flags |= TDF_NEEDRESCHED; 906 } 907 908 void 909 sched_sleep(struct thread *td, int pri) 910 { 911 912 THREAD_LOCK_ASSERT(td, MA_OWNED); 913 td->td_slptick = ticks; 914 td->td_sched->ts_slptime = 0; 915 if (pri != 0 && PRI_BASE(td->td_pri_class) == PRI_TIMESHARE) 916 sched_prio(td, pri); 917 if (TD_IS_SUSPENDED(td) || pri >= PSOCK) 918 td->td_flags |= TDF_CANSWAP; 919 } 920 921 void 922 sched_switch(struct thread *td, struct thread *newtd, int flags) 923 { 924 struct mtx *tmtx; 925 struct td_sched *ts; 926 struct proc *p; 927 928 tmtx = NULL; 929 ts = td->td_sched; 930 p = td->td_proc; 931 932 THREAD_LOCK_ASSERT(td, MA_OWNED); 933 934 /* 935 * Switch to the sched lock to fix things up and pick 936 * a new thread. 937 * Block the td_lock in order to avoid breaking the critical path. 938 */ 939 if (td->td_lock != &sched_lock) { 940 mtx_lock_spin(&sched_lock); 941 tmtx = thread_lock_block(td); 942 } 943 944 if ((td->td_flags & TDF_NOLOAD) == 0) 945 sched_load_rem(); 946 947 td->td_lastcpu = td->td_oncpu; 948 if (!(flags & SW_PREEMPT)) 949 td->td_flags &= ~TDF_NEEDRESCHED; 950 td->td_owepreempt = 0; 951 td->td_oncpu = NOCPU; 952 953 /* 954 * At the last moment, if this thread is still marked RUNNING, 955 * then put it back on the run queue as it has not been suspended 956 * or stopped or any thing else similar. We never put the idle 957 * threads on the run queue, however. 958 */ 959 if (td->td_flags & TDF_IDLETD) { 960 TD_SET_CAN_RUN(td); 961 #ifdef SMP 962 idle_cpus_mask &= ~PCPU_GET(cpumask); 963 #endif 964 } else { 965 if (TD_IS_RUNNING(td)) { 966 /* Put us back on the run queue. */ 967 sched_add(td, (flags & SW_PREEMPT) ? 968 SRQ_OURSELF|SRQ_YIELDING|SRQ_PREEMPTED : 969 SRQ_OURSELF|SRQ_YIELDING); 970 } 971 } 972 if (newtd) { 973 /* 974 * The thread we are about to run needs to be counted 975 * as if it had been added to the run queue and selected. 976 * It came from: 977 * * A preemption 978 * * An upcall 979 * * A followon 980 */ 981 KASSERT((newtd->td_inhibitors == 0), 982 ("trying to run inhibited thread")); 983 newtd->td_flags |= TDF_DIDRUN; 984 TD_SET_RUNNING(newtd); 985 if ((newtd->td_flags & TDF_NOLOAD) == 0) 986 sched_load_add(); 987 } else { 988 newtd = choosethread(); 989 MPASS(newtd->td_lock == &sched_lock); 990 } 991 992 if (td != newtd) { 993 #ifdef HWPMC_HOOKS 994 if (PMC_PROC_IS_USING_PMCS(td->td_proc)) 995 PMC_SWITCH_CONTEXT(td, PMC_FN_CSW_OUT); 996 #endif 997 /* I feel sleepy */ 998 lock_profile_release_lock(&sched_lock.lock_object); 999 #ifdef KDTRACE_HOOKS 1000 /* 1001 * If DTrace has set the active vtime enum to anything 1002 * other than INACTIVE (0), then it should have set the 1003 * function to call. 1004 */ 1005 if (dtrace_vtime_active) 1006 (*dtrace_vtime_switch_func)(newtd); 1007 #endif 1008 1009 cpu_switch(td, newtd, tmtx != NULL ? tmtx : td->td_lock); 1010 lock_profile_obtain_lock_success(&sched_lock.lock_object, 1011 0, 0, __FILE__, __LINE__); 1012 /* 1013 * Where am I? What year is it? 1014 * We are in the same thread that went to sleep above, 1015 * but any amount of time may have passed. All our context 1016 * will still be available as will local variables. 1017 * PCPU values however may have changed as we may have 1018 * changed CPU so don't trust cached values of them. 1019 * New threads will go to fork_exit() instead of here 1020 * so if you change things here you may need to change 1021 * things there too. 1022 * 1023 * If the thread above was exiting it will never wake 1024 * up again here, so either it has saved everything it 1025 * needed to, or the thread_wait() or wait() will 1026 * need to reap it. 1027 */ 1028 #ifdef HWPMC_HOOKS 1029 if (PMC_PROC_IS_USING_PMCS(td->td_proc)) 1030 PMC_SWITCH_CONTEXT(td, PMC_FN_CSW_IN); 1031 #endif 1032 } 1033 1034 #ifdef SMP 1035 if (td->td_flags & TDF_IDLETD) 1036 idle_cpus_mask |= PCPU_GET(cpumask); 1037 #endif 1038 sched_lock.mtx_lock = (uintptr_t)td; 1039 td->td_oncpu = PCPU_GET(cpuid); 1040 MPASS(td->td_lock == &sched_lock); 1041 } 1042 1043 void 1044 sched_wakeup(struct thread *td) 1045 { 1046 struct td_sched *ts; 1047 1048 THREAD_LOCK_ASSERT(td, MA_OWNED); 1049 ts = td->td_sched; 1050 td->td_flags &= ~TDF_CANSWAP; 1051 if (ts->ts_slptime > 1) { 1052 updatepri(td); 1053 resetpriority(td); 1054 } 1055 td->td_slptick = 0; 1056 ts->ts_slptime = 0; 1057 sched_add(td, SRQ_BORING); 1058 } 1059 1060 #ifdef SMP 1061 static int 1062 forward_wakeup(int cpunum) 1063 { 1064 struct pcpu *pc; 1065 cpumask_t dontuse, id, map, map2, map3, me; 1066 1067 mtx_assert(&sched_lock, MA_OWNED); 1068 1069 CTR0(KTR_RUNQ, "forward_wakeup()"); 1070 1071 if ((!forward_wakeup_enabled) || 1072 (forward_wakeup_use_mask == 0 && forward_wakeup_use_loop == 0)) 1073 return (0); 1074 if (!smp_started || cold || panicstr) 1075 return (0); 1076 1077 forward_wakeups_requested++; 1078 1079 /* 1080 * Check the idle mask we received against what we calculated 1081 * before in the old version. 1082 */ 1083 me = PCPU_GET(cpumask); 1084 1085 /* Don't bother if we should be doing it ourself. */ 1086 if ((me & idle_cpus_mask) && (cpunum == NOCPU || me == (1 << cpunum))) 1087 return (0); 1088 1089 dontuse = me | stopped_cpus | hlt_cpus_mask; 1090 map3 = 0; 1091 if (forward_wakeup_use_loop) { 1092 SLIST_FOREACH(pc, &cpuhead, pc_allcpu) { 1093 id = pc->pc_cpumask; 1094 if ((id & dontuse) == 0 && 1095 pc->pc_curthread == pc->pc_idlethread) { 1096 map3 |= id; 1097 } 1098 } 1099 } 1100 1101 if (forward_wakeup_use_mask) { 1102 map = 0; 1103 map = idle_cpus_mask & ~dontuse; 1104 1105 /* If they are both on, compare and use loop if different. */ 1106 if (forward_wakeup_use_loop) { 1107 if (map != map3) { 1108 printf("map (%02X) != map3 (%02X)\n", map, 1109 map3); 1110 map = map3; 1111 } 1112 } 1113 } else { 1114 map = map3; 1115 } 1116 1117 /* If we only allow a specific CPU, then mask off all the others. */ 1118 if (cpunum != NOCPU) { 1119 KASSERT((cpunum <= mp_maxcpus),("forward_wakeup: bad cpunum.")); 1120 map &= (1 << cpunum); 1121 } else { 1122 /* Try choose an idle die. */ 1123 if (forward_wakeup_use_htt) { 1124 map2 = (map & (map >> 1)) & 0x5555; 1125 if (map2) { 1126 map = map2; 1127 } 1128 } 1129 1130 /* Set only one bit. */ 1131 if (forward_wakeup_use_single) { 1132 map = map & ((~map) + 1); 1133 } 1134 } 1135 if (map) { 1136 forward_wakeups_delivered++; 1137 SLIST_FOREACH(pc, &cpuhead, pc_allcpu) { 1138 id = pc->pc_cpumask; 1139 if ((map & id) == 0) 1140 continue; 1141 if (cpu_idle_wakeup(pc->pc_cpuid)) 1142 map &= ~id; 1143 } 1144 if (map) 1145 ipi_selected(map, IPI_AST); 1146 return (1); 1147 } 1148 if (cpunum == NOCPU) 1149 printf("forward_wakeup: Idle processor not found\n"); 1150 return (0); 1151 } 1152 1153 static void 1154 kick_other_cpu(int pri, int cpuid) 1155 { 1156 struct pcpu *pcpu; 1157 int cpri; 1158 1159 pcpu = pcpu_find(cpuid); 1160 if (idle_cpus_mask & pcpu->pc_cpumask) { 1161 forward_wakeups_delivered++; 1162 if (!cpu_idle_wakeup(cpuid)) 1163 ipi_cpu(cpuid, IPI_AST); 1164 return; 1165 } 1166 1167 cpri = pcpu->pc_curthread->td_priority; 1168 if (pri >= cpri) 1169 return; 1170 1171 #if defined(IPI_PREEMPTION) && defined(PREEMPTION) 1172 #if !defined(FULL_PREEMPTION) 1173 if (pri <= PRI_MAX_ITHD) 1174 #endif /* ! FULL_PREEMPTION */ 1175 { 1176 ipi_cpu(cpuid, IPI_PREEMPT); 1177 return; 1178 } 1179 #endif /* defined(IPI_PREEMPTION) && defined(PREEMPTION) */ 1180 1181 pcpu->pc_curthread->td_flags |= TDF_NEEDRESCHED; 1182 ipi_cpu(cpuid, IPI_AST); 1183 return; 1184 } 1185 #endif /* SMP */ 1186 1187 #ifdef SMP 1188 static int 1189 sched_pickcpu(struct thread *td) 1190 { 1191 int best, cpu; 1192 1193 mtx_assert(&sched_lock, MA_OWNED); 1194 1195 if (THREAD_CAN_SCHED(td, td->td_lastcpu)) 1196 best = td->td_lastcpu; 1197 else 1198 best = NOCPU; 1199 CPU_FOREACH(cpu) { 1200 if (!THREAD_CAN_SCHED(td, cpu)) 1201 continue; 1202 1203 if (best == NOCPU) 1204 best = cpu; 1205 else if (runq_length[cpu] < runq_length[best]) 1206 best = cpu; 1207 } 1208 KASSERT(best != NOCPU, ("no valid CPUs")); 1209 1210 return (best); 1211 } 1212 #endif 1213 1214 void 1215 sched_add(struct thread *td, int flags) 1216 #ifdef SMP 1217 { 1218 struct td_sched *ts; 1219 int forwarded = 0; 1220 int cpu; 1221 int single_cpu = 0; 1222 1223 ts = td->td_sched; 1224 THREAD_LOCK_ASSERT(td, MA_OWNED); 1225 KASSERT((td->td_inhibitors == 0), 1226 ("sched_add: trying to run inhibited thread")); 1227 KASSERT((TD_CAN_RUN(td) || TD_IS_RUNNING(td)), 1228 ("sched_add: bad thread state")); 1229 KASSERT(td->td_flags & TDF_INMEM, 1230 ("sched_add: thread swapped out")); 1231 1232 KTR_STATE2(KTR_SCHED, "thread", sched_tdname(td), "runq add", 1233 "prio:%d", td->td_priority, KTR_ATTR_LINKED, 1234 sched_tdname(curthread)); 1235 KTR_POINT1(KTR_SCHED, "thread", sched_tdname(curthread), "wokeup", 1236 KTR_ATTR_LINKED, sched_tdname(td)); 1237 1238 1239 /* 1240 * Now that the thread is moving to the run-queue, set the lock 1241 * to the scheduler's lock. 1242 */ 1243 if (td->td_lock != &sched_lock) { 1244 mtx_lock_spin(&sched_lock); 1245 thread_lock_set(td, &sched_lock); 1246 } 1247 TD_SET_RUNQ(td); 1248 1249 if (td->td_pinned != 0) { 1250 cpu = td->td_lastcpu; 1251 ts->ts_runq = &runq_pcpu[cpu]; 1252 single_cpu = 1; 1253 CTR3(KTR_RUNQ, 1254 "sched_add: Put td_sched:%p(td:%p) on cpu%d runq", ts, td, 1255 cpu); 1256 } else if (td->td_flags & TDF_BOUND) { 1257 /* Find CPU from bound runq. */ 1258 KASSERT(SKE_RUNQ_PCPU(ts), 1259 ("sched_add: bound td_sched not on cpu runq")); 1260 cpu = ts->ts_runq - &runq_pcpu[0]; 1261 single_cpu = 1; 1262 CTR3(KTR_RUNQ, 1263 "sched_add: Put td_sched:%p(td:%p) on cpu%d runq", ts, td, 1264 cpu); 1265 } else if (ts->ts_flags & TSF_AFFINITY) { 1266 /* Find a valid CPU for our cpuset */ 1267 cpu = sched_pickcpu(td); 1268 ts->ts_runq = &runq_pcpu[cpu]; 1269 single_cpu = 1; 1270 CTR3(KTR_RUNQ, 1271 "sched_add: Put td_sched:%p(td:%p) on cpu%d runq", ts, td, 1272 cpu); 1273 } else { 1274 CTR2(KTR_RUNQ, 1275 "sched_add: adding td_sched:%p (td:%p) to gbl runq", ts, 1276 td); 1277 cpu = NOCPU; 1278 ts->ts_runq = &runq; 1279 } 1280 1281 if (single_cpu && (cpu != PCPU_GET(cpuid))) { 1282 kick_other_cpu(td->td_priority, cpu); 1283 } else { 1284 if (!single_cpu) { 1285 cpumask_t me = PCPU_GET(cpumask); 1286 cpumask_t idle = idle_cpus_mask & me; 1287 1288 if (!idle && ((flags & SRQ_INTR) == 0) && 1289 (idle_cpus_mask & ~(hlt_cpus_mask | me))) 1290 forwarded = forward_wakeup(cpu); 1291 } 1292 1293 if (!forwarded) { 1294 if ((flags & SRQ_YIELDING) == 0 && maybe_preempt(td)) 1295 return; 1296 else 1297 maybe_resched(td); 1298 } 1299 } 1300 1301 if ((td->td_flags & TDF_NOLOAD) == 0) 1302 sched_load_add(); 1303 runq_add(ts->ts_runq, td, flags); 1304 if (cpu != NOCPU) 1305 runq_length[cpu]++; 1306 } 1307 #else /* SMP */ 1308 { 1309 struct td_sched *ts; 1310 1311 ts = td->td_sched; 1312 THREAD_LOCK_ASSERT(td, MA_OWNED); 1313 KASSERT((td->td_inhibitors == 0), 1314 ("sched_add: trying to run inhibited thread")); 1315 KASSERT((TD_CAN_RUN(td) || TD_IS_RUNNING(td)), 1316 ("sched_add: bad thread state")); 1317 KASSERT(td->td_flags & TDF_INMEM, 1318 ("sched_add: thread swapped out")); 1319 KTR_STATE2(KTR_SCHED, "thread", sched_tdname(td), "runq add", 1320 "prio:%d", td->td_priority, KTR_ATTR_LINKED, 1321 sched_tdname(curthread)); 1322 KTR_POINT1(KTR_SCHED, "thread", sched_tdname(curthread), "wokeup", 1323 KTR_ATTR_LINKED, sched_tdname(td)); 1324 1325 /* 1326 * Now that the thread is moving to the run-queue, set the lock 1327 * to the scheduler's lock. 1328 */ 1329 if (td->td_lock != &sched_lock) { 1330 mtx_lock_spin(&sched_lock); 1331 thread_lock_set(td, &sched_lock); 1332 } 1333 TD_SET_RUNQ(td); 1334 CTR2(KTR_RUNQ, "sched_add: adding td_sched:%p (td:%p) to runq", ts, td); 1335 ts->ts_runq = &runq; 1336 1337 /* 1338 * If we are yielding (on the way out anyhow) or the thread 1339 * being saved is US, then don't try be smart about preemption 1340 * or kicking off another CPU as it won't help and may hinder. 1341 * In the YIEDLING case, we are about to run whoever is being 1342 * put in the queue anyhow, and in the OURSELF case, we are 1343 * puting ourself on the run queue which also only happens 1344 * when we are about to yield. 1345 */ 1346 if ((flags & SRQ_YIELDING) == 0) { 1347 if (maybe_preempt(td)) 1348 return; 1349 } 1350 if ((td->td_flags & TDF_NOLOAD) == 0) 1351 sched_load_add(); 1352 runq_add(ts->ts_runq, td, flags); 1353 maybe_resched(td); 1354 } 1355 #endif /* SMP */ 1356 1357 void 1358 sched_rem(struct thread *td) 1359 { 1360 struct td_sched *ts; 1361 1362 ts = td->td_sched; 1363 KASSERT(td->td_flags & TDF_INMEM, 1364 ("sched_rem: thread swapped out")); 1365 KASSERT(TD_ON_RUNQ(td), 1366 ("sched_rem: thread not on run queue")); 1367 mtx_assert(&sched_lock, MA_OWNED); 1368 KTR_STATE2(KTR_SCHED, "thread", sched_tdname(td), "runq rem", 1369 "prio:%d", td->td_priority, KTR_ATTR_LINKED, 1370 sched_tdname(curthread)); 1371 1372 if ((td->td_flags & TDF_NOLOAD) == 0) 1373 sched_load_rem(); 1374 #ifdef SMP 1375 if (ts->ts_runq != &runq) 1376 runq_length[ts->ts_runq - runq_pcpu]--; 1377 #endif 1378 runq_remove(ts->ts_runq, td); 1379 TD_SET_CAN_RUN(td); 1380 } 1381 1382 /* 1383 * Select threads to run. Note that running threads still consume a 1384 * slot. 1385 */ 1386 struct thread * 1387 sched_choose(void) 1388 { 1389 struct thread *td; 1390 struct runq *rq; 1391 1392 mtx_assert(&sched_lock, MA_OWNED); 1393 #ifdef SMP 1394 struct thread *tdcpu; 1395 1396 rq = &runq; 1397 td = runq_choose_fuzz(&runq, runq_fuzz); 1398 tdcpu = runq_choose(&runq_pcpu[PCPU_GET(cpuid)]); 1399 1400 if (td == NULL || 1401 (tdcpu != NULL && 1402 tdcpu->td_priority < td->td_priority)) { 1403 CTR2(KTR_RUNQ, "choosing td %p from pcpu runq %d", tdcpu, 1404 PCPU_GET(cpuid)); 1405 td = tdcpu; 1406 rq = &runq_pcpu[PCPU_GET(cpuid)]; 1407 } else { 1408 CTR1(KTR_RUNQ, "choosing td_sched %p from main runq", td); 1409 } 1410 1411 #else 1412 rq = &runq; 1413 td = runq_choose(&runq); 1414 #endif 1415 1416 if (td) { 1417 #ifdef SMP 1418 if (td == tdcpu) 1419 runq_length[PCPU_GET(cpuid)]--; 1420 #endif 1421 runq_remove(rq, td); 1422 td->td_flags |= TDF_DIDRUN; 1423 1424 KASSERT(td->td_flags & TDF_INMEM, 1425 ("sched_choose: thread swapped out")); 1426 return (td); 1427 } 1428 return (PCPU_GET(idlethread)); 1429 } 1430 1431 void 1432 sched_preempt(struct thread *td) 1433 { 1434 thread_lock(td); 1435 if (td->td_critnest > 1) 1436 td->td_owepreempt = 1; 1437 else 1438 mi_switch(SW_INVOL | SW_PREEMPT | SWT_PREEMPT, NULL); 1439 thread_unlock(td); 1440 } 1441 1442 void 1443 sched_userret(struct thread *td) 1444 { 1445 /* 1446 * XXX we cheat slightly on the locking here to avoid locking in 1447 * the usual case. Setting td_priority here is essentially an 1448 * incomplete workaround for not setting it properly elsewhere. 1449 * Now that some interrupt handlers are threads, not setting it 1450 * properly elsewhere can clobber it in the window between setting 1451 * it here and returning to user mode, so don't waste time setting 1452 * it perfectly here. 1453 */ 1454 KASSERT((td->td_flags & TDF_BORROWING) == 0, 1455 ("thread with borrowed priority returning to userland")); 1456 if (td->td_priority != td->td_user_pri) { 1457 thread_lock(td); 1458 td->td_priority = td->td_user_pri; 1459 td->td_base_pri = td->td_user_pri; 1460 thread_unlock(td); 1461 } 1462 } 1463 1464 void 1465 sched_bind(struct thread *td, int cpu) 1466 { 1467 struct td_sched *ts; 1468 1469 THREAD_LOCK_ASSERT(td, MA_OWNED|MA_NOTRECURSED); 1470 KASSERT(td == curthread, ("sched_bind: can only bind curthread")); 1471 1472 ts = td->td_sched; 1473 1474 td->td_flags |= TDF_BOUND; 1475 #ifdef SMP 1476 ts->ts_runq = &runq_pcpu[cpu]; 1477 if (PCPU_GET(cpuid) == cpu) 1478 return; 1479 1480 mi_switch(SW_VOL, NULL); 1481 #endif 1482 } 1483 1484 void 1485 sched_unbind(struct thread* td) 1486 { 1487 THREAD_LOCK_ASSERT(td, MA_OWNED); 1488 KASSERT(td == curthread, ("sched_unbind: can only bind curthread")); 1489 td->td_flags &= ~TDF_BOUND; 1490 } 1491 1492 int 1493 sched_is_bound(struct thread *td) 1494 { 1495 THREAD_LOCK_ASSERT(td, MA_OWNED); 1496 return (td->td_flags & TDF_BOUND); 1497 } 1498 1499 void 1500 sched_relinquish(struct thread *td) 1501 { 1502 thread_lock(td); 1503 mi_switch(SW_VOL | SWT_RELINQUISH, NULL); 1504 thread_unlock(td); 1505 } 1506 1507 int 1508 sched_load(void) 1509 { 1510 return (sched_tdcnt); 1511 } 1512 1513 int 1514 sched_sizeof_proc(void) 1515 { 1516 return (sizeof(struct proc)); 1517 } 1518 1519 int 1520 sched_sizeof_thread(void) 1521 { 1522 return (sizeof(struct thread) + sizeof(struct td_sched)); 1523 } 1524 1525 fixpt_t 1526 sched_pctcpu(struct thread *td) 1527 { 1528 struct td_sched *ts; 1529 1530 THREAD_LOCK_ASSERT(td, MA_OWNED); 1531 ts = td->td_sched; 1532 return (ts->ts_pctcpu); 1533 } 1534 1535 void 1536 sched_tick(int cnt) 1537 { 1538 } 1539 1540 /* 1541 * The actual idle process. 1542 */ 1543 void 1544 sched_idletd(void *dummy) 1545 { 1546 struct pcpuidlestat *stat; 1547 1548 stat = DPCPU_PTR(idlestat); 1549 for (;;) { 1550 mtx_assert(&Giant, MA_NOTOWNED); 1551 1552 while (sched_runnable() == 0) { 1553 cpu_idle(stat->idlecalls + stat->oldidlecalls > 64); 1554 stat->idlecalls++; 1555 } 1556 1557 mtx_lock_spin(&sched_lock); 1558 mi_switch(SW_VOL | SWT_IDLE, NULL); 1559 mtx_unlock_spin(&sched_lock); 1560 } 1561 } 1562 1563 /* 1564 * A CPU is entering for the first time or a thread is exiting. 1565 */ 1566 void 1567 sched_throw(struct thread *td) 1568 { 1569 /* 1570 * Correct spinlock nesting. The idle thread context that we are 1571 * borrowing was created so that it would start out with a single 1572 * spin lock (sched_lock) held in fork_trampoline(). Since we've 1573 * explicitly acquired locks in this function, the nesting count 1574 * is now 2 rather than 1. Since we are nested, calling 1575 * spinlock_exit() will simply adjust the counts without allowing 1576 * spin lock using code to interrupt us. 1577 */ 1578 if (td == NULL) { 1579 mtx_lock_spin(&sched_lock); 1580 spinlock_exit(); 1581 } else { 1582 lock_profile_release_lock(&sched_lock.lock_object); 1583 MPASS(td->td_lock == &sched_lock); 1584 } 1585 mtx_assert(&sched_lock, MA_OWNED); 1586 KASSERT(curthread->td_md.md_spinlock_count == 1, ("invalid count")); 1587 PCPU_SET(switchtime, cpu_ticks()); 1588 PCPU_SET(switchticks, ticks); 1589 cpu_throw(td, choosethread()); /* doesn't return */ 1590 } 1591 1592 void 1593 sched_fork_exit(struct thread *td) 1594 { 1595 1596 /* 1597 * Finish setting up thread glue so that it begins execution in a 1598 * non-nested critical section with sched_lock held but not recursed. 1599 */ 1600 td->td_oncpu = PCPU_GET(cpuid); 1601 sched_lock.mtx_lock = (uintptr_t)td; 1602 lock_profile_obtain_lock_success(&sched_lock.lock_object, 1603 0, 0, __FILE__, __LINE__); 1604 THREAD_LOCK_ASSERT(td, MA_OWNED | MA_NOTRECURSED); 1605 } 1606 1607 char * 1608 sched_tdname(struct thread *td) 1609 { 1610 #ifdef KTR 1611 struct td_sched *ts; 1612 1613 ts = td->td_sched; 1614 if (ts->ts_name[0] == '\0') 1615 snprintf(ts->ts_name, sizeof(ts->ts_name), 1616 "%s tid %d", td->td_name, td->td_tid); 1617 return (ts->ts_name); 1618 #else 1619 return (td->td_name); 1620 #endif 1621 } 1622 1623 void 1624 sched_affinity(struct thread *td) 1625 { 1626 #ifdef SMP 1627 struct td_sched *ts; 1628 int cpu; 1629 1630 THREAD_LOCK_ASSERT(td, MA_OWNED); 1631 1632 /* 1633 * Set the TSF_AFFINITY flag if there is at least one CPU this 1634 * thread can't run on. 1635 */ 1636 ts = td->td_sched; 1637 ts->ts_flags &= ~TSF_AFFINITY; 1638 CPU_FOREACH(cpu) { 1639 if (!THREAD_CAN_SCHED(td, cpu)) { 1640 ts->ts_flags |= TSF_AFFINITY; 1641 break; 1642 } 1643 } 1644 1645 /* 1646 * If this thread can run on all CPUs, nothing else to do. 1647 */ 1648 if (!(ts->ts_flags & TSF_AFFINITY)) 1649 return; 1650 1651 /* Pinned threads and bound threads should be left alone. */ 1652 if (td->td_pinned != 0 || td->td_flags & TDF_BOUND) 1653 return; 1654 1655 switch (td->td_state) { 1656 case TDS_RUNQ: 1657 /* 1658 * If we are on a per-CPU runqueue that is in the set, 1659 * then nothing needs to be done. 1660 */ 1661 if (ts->ts_runq != &runq && 1662 THREAD_CAN_SCHED(td, ts->ts_runq - runq_pcpu)) 1663 return; 1664 1665 /* Put this thread on a valid per-CPU runqueue. */ 1666 sched_rem(td); 1667 sched_add(td, SRQ_BORING); 1668 break; 1669 case TDS_RUNNING: 1670 /* 1671 * See if our current CPU is in the set. If not, force a 1672 * context switch. 1673 */ 1674 if (THREAD_CAN_SCHED(td, td->td_oncpu)) 1675 return; 1676 1677 td->td_flags |= TDF_NEEDRESCHED; 1678 if (td != curthread) 1679 ipi_cpu(cpu, IPI_AST); 1680 break; 1681 default: 1682 break; 1683 } 1684 #endif 1685 } 1686