xref: /freebsd/sys/kern/kern_switch.c (revision cbd30a72ca196976c1c700400ecd424baa1b9c16)
1 /*-
2  * Copyright (c) 2001 Jake Burkholder <jake@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, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  *
14  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24  * SUCH DAMAGE.
25  */
26 
27 
28 #include <sys/cdefs.h>
29 __FBSDID("$FreeBSD$");
30 
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/queue.h>
42 #include <sys/sched.h>
43 #include <sys/smp.h>
44 #include <sys/sysctl.h>
45 
46 #include <machine/cpu.h>
47 
48 /* Uncomment this to enable logging of critical_enter/exit. */
49 #if 0
50 #define	KTR_CRITICAL	KTR_SCHED
51 #else
52 #define	KTR_CRITICAL	0
53 #endif
54 
55 #ifdef FULL_PREEMPTION
56 #ifndef PREEMPTION
57 #error "The FULL_PREEMPTION option requires the PREEMPTION option"
58 #endif
59 #endif
60 
61 CTASSERT((RQB_BPW * RQB_LEN) == RQ_NQS);
62 
63 /*
64  * kern.sched.preemption allows user space to determine if preemption support
65  * is compiled in or not.  It is not currently a boot or runtime flag that
66  * can be changed.
67  */
68 #ifdef PREEMPTION
69 static int kern_sched_preemption = 1;
70 #else
71 static int kern_sched_preemption = 0;
72 #endif
73 SYSCTL_INT(_kern_sched, OID_AUTO, preemption, CTLFLAG_RD,
74     &kern_sched_preemption, 0, "Kernel preemption enabled");
75 
76 /*
77  * Support for scheduler stats exported via kern.sched.stats.  All stats may
78  * be reset with kern.sched.stats.reset = 1.  Stats may be defined elsewhere
79  * with SCHED_STAT_DEFINE().
80  */
81 #ifdef SCHED_STATS
82 SYSCTL_NODE(_kern_sched, OID_AUTO, stats, CTLFLAG_RW, 0, "switch stats");
83 
84 /* Switch reasons from mi_switch(). */
85 DPCPU_DEFINE(long, sched_switch_stats[SWT_COUNT]);
86 SCHED_STAT_DEFINE_VAR(uncategorized,
87     &DPCPU_NAME(sched_switch_stats[SWT_NONE]), "");
88 SCHED_STAT_DEFINE_VAR(preempt,
89     &DPCPU_NAME(sched_switch_stats[SWT_PREEMPT]), "");
90 SCHED_STAT_DEFINE_VAR(owepreempt,
91     &DPCPU_NAME(sched_switch_stats[SWT_OWEPREEMPT]), "");
92 SCHED_STAT_DEFINE_VAR(turnstile,
93     &DPCPU_NAME(sched_switch_stats[SWT_TURNSTILE]), "");
94 SCHED_STAT_DEFINE_VAR(sleepq,
95     &DPCPU_NAME(sched_switch_stats[SWT_SLEEPQ]), "");
96 SCHED_STAT_DEFINE_VAR(sleepqtimo,
97     &DPCPU_NAME(sched_switch_stats[SWT_SLEEPQTIMO]), "");
98 SCHED_STAT_DEFINE_VAR(relinquish,
99     &DPCPU_NAME(sched_switch_stats[SWT_RELINQUISH]), "");
100 SCHED_STAT_DEFINE_VAR(needresched,
101     &DPCPU_NAME(sched_switch_stats[SWT_NEEDRESCHED]), "");
102 SCHED_STAT_DEFINE_VAR(idle,
103     &DPCPU_NAME(sched_switch_stats[SWT_IDLE]), "");
104 SCHED_STAT_DEFINE_VAR(iwait,
105     &DPCPU_NAME(sched_switch_stats[SWT_IWAIT]), "");
106 SCHED_STAT_DEFINE_VAR(suspend,
107     &DPCPU_NAME(sched_switch_stats[SWT_SUSPEND]), "");
108 SCHED_STAT_DEFINE_VAR(remotepreempt,
109     &DPCPU_NAME(sched_switch_stats[SWT_REMOTEPREEMPT]), "");
110 SCHED_STAT_DEFINE_VAR(remotewakeidle,
111     &DPCPU_NAME(sched_switch_stats[SWT_REMOTEWAKEIDLE]), "");
112 
113 static int
114 sysctl_stats_reset(SYSCTL_HANDLER_ARGS)
115 {
116 	struct sysctl_oid *p;
117 	uintptr_t counter;
118         int error;
119 	int val;
120 	int i;
121 
122         val = 0;
123         error = sysctl_handle_int(oidp, &val, 0, req);
124         if (error != 0 || req->newptr == NULL)
125                 return (error);
126         if (val == 0)
127                 return (0);
128 	/*
129 	 * Traverse the list of children of _kern_sched_stats and reset each
130 	 * to 0.  Skip the reset entry.
131 	 */
132 	SLIST_FOREACH(p, oidp->oid_parent, oid_link) {
133 		if (p == oidp || p->oid_arg1 == NULL)
134 			continue;
135 		counter = (uintptr_t)p->oid_arg1;
136 		CPU_FOREACH(i) {
137 			*(long *)(dpcpu_off[i] + counter) = 0;
138 		}
139 	}
140 	return (0);
141 }
142 
143 SYSCTL_PROC(_kern_sched_stats, OID_AUTO, reset, CTLTYPE_INT | CTLFLAG_WR, NULL,
144     0, sysctl_stats_reset, "I", "Reset scheduler statistics");
145 #endif
146 
147 /************************************************************************
148  * Functions that manipulate runnability from a thread perspective.	*
149  ************************************************************************/
150 /*
151  * Select the thread that will be run next.
152  */
153 struct thread *
154 choosethread(void)
155 {
156 	struct thread *td;
157 
158 retry:
159 	td = sched_choose();
160 
161 	/*
162 	 * If we are in panic, only allow system threads,
163 	 * plus the one we are running in, to be run.
164 	 */
165 	if (panicstr && ((td->td_proc->p_flag & P_SYSTEM) == 0 &&
166 	    (td->td_flags & TDF_INPANIC) == 0)) {
167 		/* note that it is no longer on the run queue */
168 		TD_SET_CAN_RUN(td);
169 		goto retry;
170 	}
171 
172 	TD_SET_RUNNING(td);
173 	return (td);
174 }
175 
176 /*
177  * Kernel thread preemption implementation.  Critical sections mark
178  * regions of code in which preemptions are not allowed.
179  *
180  * It might seem a good idea to inline critical_enter() but, in order
181  * to prevent instructions reordering by the compiler, a __compiler_membar()
182  * would have to be used here (the same as sched_pin()).  The performance
183  * penalty imposed by the membar could, then, produce slower code than
184  * the function call itself, for most cases.
185  */
186 void
187 critical_enter(void)
188 {
189 	struct thread *td;
190 
191 	td = curthread;
192 	td->td_critnest++;
193 	CTR4(KTR_CRITICAL, "critical_enter by thread %p (%ld, %s) to %d", td,
194 	    (long)td->td_proc->p_pid, td->td_name, td->td_critnest);
195 }
196 
197 void
198 critical_exit(void)
199 {
200 	struct thread *td;
201 	int flags;
202 
203 	td = curthread;
204 	KASSERT(td->td_critnest != 0,
205 	    ("critical_exit: td_critnest == 0"));
206 
207 	if (td->td_critnest == 1) {
208 		td->td_critnest = 0;
209 
210 		/*
211 		 * Interrupt handlers execute critical_exit() on
212 		 * leave, and td_owepreempt may be left set by an
213 		 * interrupt handler only when td_critnest > 0.  If we
214 		 * are decrementing td_critnest from 1 to 0, read
215 		 * td_owepreempt after decrementing, to not miss the
216 		 * preempt.  Disallow compiler to reorder operations.
217 		 */
218 		__compiler_membar();
219 		if (td->td_owepreempt && !kdb_active) {
220 			/*
221 			 * Microoptimization: we committed to switch,
222 			 * disable preemption in interrupt handlers
223 			 * while spinning for the thread lock.
224 			 */
225 			td->td_critnest = 1;
226 			thread_lock(td);
227 			td->td_critnest--;
228 			flags = SW_INVOL | SW_PREEMPT;
229 			if (TD_IS_IDLETHREAD(td))
230 				flags |= SWT_IDLE;
231 			else
232 				flags |= SWT_OWEPREEMPT;
233 			mi_switch(flags, NULL);
234 			thread_unlock(td);
235 		}
236 	} else
237 		td->td_critnest--;
238 
239 	CTR4(KTR_CRITICAL, "critical_exit by thread %p (%ld, %s) to %d", td,
240 	    (long)td->td_proc->p_pid, td->td_name, td->td_critnest);
241 }
242 
243 /************************************************************************
244  * SYSTEM RUN QUEUE manipulations and tests				*
245  ************************************************************************/
246 /*
247  * Initialize a run structure.
248  */
249 void
250 runq_init(struct runq *rq)
251 {
252 	int i;
253 
254 	bzero(rq, sizeof *rq);
255 	for (i = 0; i < RQ_NQS; i++)
256 		TAILQ_INIT(&rq->rq_queues[i]);
257 }
258 
259 /*
260  * Clear the status bit of the queue corresponding to priority level pri,
261  * indicating that it is empty.
262  */
263 static __inline void
264 runq_clrbit(struct runq *rq, int pri)
265 {
266 	struct rqbits *rqb;
267 
268 	rqb = &rq->rq_status;
269 	CTR4(KTR_RUNQ, "runq_clrbit: bits=%#x %#x bit=%#x word=%d",
270 	    rqb->rqb_bits[RQB_WORD(pri)],
271 	    rqb->rqb_bits[RQB_WORD(pri)] & ~RQB_BIT(pri),
272 	    RQB_BIT(pri), RQB_WORD(pri));
273 	rqb->rqb_bits[RQB_WORD(pri)] &= ~RQB_BIT(pri);
274 }
275 
276 /*
277  * Find the index of the first non-empty run queue.  This is done by
278  * scanning the status bits, a set bit indicates a non-empty queue.
279  */
280 static __inline int
281 runq_findbit(struct runq *rq)
282 {
283 	struct rqbits *rqb;
284 	int pri;
285 	int i;
286 
287 	rqb = &rq->rq_status;
288 	for (i = 0; i < RQB_LEN; i++)
289 		if (rqb->rqb_bits[i]) {
290 			pri = RQB_FFS(rqb->rqb_bits[i]) + (i << RQB_L2BPW);
291 			CTR3(KTR_RUNQ, "runq_findbit: bits=%#x i=%d pri=%d",
292 			    rqb->rqb_bits[i], i, pri);
293 			return (pri);
294 		}
295 
296 	return (-1);
297 }
298 
299 static __inline int
300 runq_findbit_from(struct runq *rq, u_char pri)
301 {
302 	struct rqbits *rqb;
303 	rqb_word_t mask;
304 	int i;
305 
306 	/*
307 	 * Set the mask for the first word so we ignore priorities before 'pri'.
308 	 */
309 	mask = (rqb_word_t)-1 << (pri & (RQB_BPW - 1));
310 	rqb = &rq->rq_status;
311 again:
312 	for (i = RQB_WORD(pri); i < RQB_LEN; mask = -1, i++) {
313 		mask = rqb->rqb_bits[i] & mask;
314 		if (mask == 0)
315 			continue;
316 		pri = RQB_FFS(mask) + (i << RQB_L2BPW);
317 		CTR3(KTR_RUNQ, "runq_findbit_from: bits=%#x i=%d pri=%d",
318 		    mask, i, pri);
319 		return (pri);
320 	}
321 	if (pri == 0)
322 		return (-1);
323 	/*
324 	 * Wrap back around to the beginning of the list just once so we
325 	 * scan the whole thing.
326 	 */
327 	pri = 0;
328 	goto again;
329 }
330 
331 /*
332  * Set the status bit of the queue corresponding to priority level pri,
333  * indicating that it is non-empty.
334  */
335 static __inline void
336 runq_setbit(struct runq *rq, int pri)
337 {
338 	struct rqbits *rqb;
339 
340 	rqb = &rq->rq_status;
341 	CTR4(KTR_RUNQ, "runq_setbit: bits=%#x %#x bit=%#x word=%d",
342 	    rqb->rqb_bits[RQB_WORD(pri)],
343 	    rqb->rqb_bits[RQB_WORD(pri)] | RQB_BIT(pri),
344 	    RQB_BIT(pri), RQB_WORD(pri));
345 	rqb->rqb_bits[RQB_WORD(pri)] |= RQB_BIT(pri);
346 }
347 
348 /*
349  * Add the thread to the queue specified by its priority, and set the
350  * corresponding status bit.
351  */
352 void
353 runq_add(struct runq *rq, struct thread *td, int flags)
354 {
355 	struct rqhead *rqh;
356 	int pri;
357 
358 	pri = td->td_priority / RQ_PPQ;
359 	td->td_rqindex = pri;
360 	runq_setbit(rq, pri);
361 	rqh = &rq->rq_queues[pri];
362 	CTR4(KTR_RUNQ, "runq_add: td=%p pri=%d %d rqh=%p",
363 	    td, td->td_priority, pri, rqh);
364 	if (flags & SRQ_PREEMPTED) {
365 		TAILQ_INSERT_HEAD(rqh, td, td_runq);
366 	} else {
367 		TAILQ_INSERT_TAIL(rqh, td, td_runq);
368 	}
369 }
370 
371 void
372 runq_add_pri(struct runq *rq, struct thread *td, u_char pri, int flags)
373 {
374 	struct rqhead *rqh;
375 
376 	KASSERT(pri < RQ_NQS, ("runq_add_pri: %d out of range", pri));
377 	td->td_rqindex = pri;
378 	runq_setbit(rq, pri);
379 	rqh = &rq->rq_queues[pri];
380 	CTR4(KTR_RUNQ, "runq_add_pri: td=%p pri=%d idx=%d rqh=%p",
381 	    td, td->td_priority, pri, rqh);
382 	if (flags & SRQ_PREEMPTED) {
383 		TAILQ_INSERT_HEAD(rqh, td, td_runq);
384 	} else {
385 		TAILQ_INSERT_TAIL(rqh, td, td_runq);
386 	}
387 }
388 /*
389  * Return true if there are runnable processes of any priority on the run
390  * queue, false otherwise.  Has no side effects, does not modify the run
391  * queue structure.
392  */
393 int
394 runq_check(struct runq *rq)
395 {
396 	struct rqbits *rqb;
397 	int i;
398 
399 	rqb = &rq->rq_status;
400 	for (i = 0; i < RQB_LEN; i++)
401 		if (rqb->rqb_bits[i]) {
402 			CTR2(KTR_RUNQ, "runq_check: bits=%#x i=%d",
403 			    rqb->rqb_bits[i], i);
404 			return (1);
405 		}
406 	CTR0(KTR_RUNQ, "runq_check: empty");
407 
408 	return (0);
409 }
410 
411 /*
412  * Find the highest priority process on the run queue.
413  */
414 struct thread *
415 runq_choose_fuzz(struct runq *rq, int fuzz)
416 {
417 	struct rqhead *rqh;
418 	struct thread *td;
419 	int pri;
420 
421 	while ((pri = runq_findbit(rq)) != -1) {
422 		rqh = &rq->rq_queues[pri];
423 		/* fuzz == 1 is normal.. 0 or less are ignored */
424 		if (fuzz > 1) {
425 			/*
426 			 * In the first couple of entries, check if
427 			 * there is one for our CPU as a preference.
428 			 */
429 			int count = fuzz;
430 			int cpu = PCPU_GET(cpuid);
431 			struct thread *td2;
432 			td2 = td = TAILQ_FIRST(rqh);
433 
434 			while (count-- && td2) {
435 				if (td2->td_lastcpu == cpu) {
436 					td = td2;
437 					break;
438 				}
439 				td2 = TAILQ_NEXT(td2, td_runq);
440 			}
441 		} else
442 			td = TAILQ_FIRST(rqh);
443 		KASSERT(td != NULL, ("runq_choose_fuzz: no proc on busy queue"));
444 		CTR3(KTR_RUNQ,
445 		    "runq_choose_fuzz: pri=%d thread=%p rqh=%p", pri, td, rqh);
446 		return (td);
447 	}
448 	CTR1(KTR_RUNQ, "runq_choose_fuzz: idleproc pri=%d", pri);
449 
450 	return (NULL);
451 }
452 
453 /*
454  * Find the highest priority process on the run queue.
455  */
456 struct thread *
457 runq_choose(struct runq *rq)
458 {
459 	struct rqhead *rqh;
460 	struct thread *td;
461 	int pri;
462 
463 	while ((pri = runq_findbit(rq)) != -1) {
464 		rqh = &rq->rq_queues[pri];
465 		td = TAILQ_FIRST(rqh);
466 		KASSERT(td != NULL, ("runq_choose: no thread on busy queue"));
467 		CTR3(KTR_RUNQ,
468 		    "runq_choose: pri=%d thread=%p rqh=%p", pri, td, rqh);
469 		return (td);
470 	}
471 	CTR1(KTR_RUNQ, "runq_choose: idlethread pri=%d", pri);
472 
473 	return (NULL);
474 }
475 
476 struct thread *
477 runq_choose_from(struct runq *rq, u_char idx)
478 {
479 	struct rqhead *rqh;
480 	struct thread *td;
481 	int pri;
482 
483 	if ((pri = runq_findbit_from(rq, idx)) != -1) {
484 		rqh = &rq->rq_queues[pri];
485 		td = TAILQ_FIRST(rqh);
486 		KASSERT(td != NULL, ("runq_choose: no thread on busy queue"));
487 		CTR4(KTR_RUNQ,
488 		    "runq_choose_from: pri=%d thread=%p idx=%d rqh=%p",
489 		    pri, td, td->td_rqindex, rqh);
490 		return (td);
491 	}
492 	CTR1(KTR_RUNQ, "runq_choose_from: idlethread pri=%d", pri);
493 
494 	return (NULL);
495 }
496 /*
497  * Remove the thread from the queue specified by its priority, and clear the
498  * corresponding status bit if the queue becomes empty.
499  * Caller must set state afterwards.
500  */
501 void
502 runq_remove(struct runq *rq, struct thread *td)
503 {
504 
505 	runq_remove_idx(rq, td, NULL);
506 }
507 
508 void
509 runq_remove_idx(struct runq *rq, struct thread *td, u_char *idx)
510 {
511 	struct rqhead *rqh;
512 	u_char pri;
513 
514 	KASSERT(td->td_flags & TDF_INMEM,
515 		("runq_remove_idx: thread swapped out"));
516 	pri = td->td_rqindex;
517 	KASSERT(pri < RQ_NQS, ("runq_remove_idx: Invalid index %d\n", pri));
518 	rqh = &rq->rq_queues[pri];
519 	CTR4(KTR_RUNQ, "runq_remove_idx: td=%p, pri=%d %d rqh=%p",
520 	    td, td->td_priority, pri, rqh);
521 	TAILQ_REMOVE(rqh, td, td_runq);
522 	if (TAILQ_EMPTY(rqh)) {
523 		CTR0(KTR_RUNQ, "runq_remove_idx: empty");
524 		runq_clrbit(rq, pri);
525 		if (idx != NULL && *idx == pri)
526 			*idx = (pri + 1) % RQ_NQS;
527 	}
528 }
529