FreeBSD/Linux Kernel Cross Reference
sys/kern/kern_synch.c
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 * @(#)kern_synch.c 8.9 (Berkeley) 5/19/95
35 */
36
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD: releng/6.1/sys/kern/kern_synch.c 158179 2006-04-30 16:44:43Z cvs2svn $");
39
40 #include "opt_ktrace.h"
41
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/condvar.h>
45 #include <sys/kdb.h>
46 #include <sys/kernel.h>
47 #include <sys/ktr.h>
48 #include <sys/lock.h>
49 #include <sys/mutex.h>
50 #include <sys/proc.h>
51 #include <sys/resourcevar.h>
52 #include <sys/sched.h>
53 #include <sys/signalvar.h>
54 #include <sys/sleepqueue.h>
55 #include <sys/smp.h>
56 #include <sys/sx.h>
57 #include <sys/sysctl.h>
58 #include <sys/sysproto.h>
59 #include <sys/vmmeter.h>
60 #ifdef KTRACE
61 #include <sys/uio.h>
62 #include <sys/ktrace.h>
63 #endif
64
65 #include <machine/cpu.h>
66
67 static void synch_setup(void *dummy);
68 SYSINIT(synch_setup, SI_SUB_KICK_SCHEDULER, SI_ORDER_FIRST, synch_setup, NULL)
69
70 int hogticks;
71 int lbolt;
72
73 static struct callout loadav_callout;
74 static struct callout lbolt_callout;
75
76 struct loadavg averunnable =
77 { {0, 0, 0}, FSCALE }; /* load average, of runnable procs */
78 /*
79 * Constants for averages over 1, 5, and 15 minutes
80 * when sampling at 5 second intervals.
81 */
82 static fixpt_t cexp[3] = {
83 0.9200444146293232 * FSCALE, /* exp(-1/12) */
84 0.9834714538216174 * FSCALE, /* exp(-1/60) */
85 0.9944598480048967 * FSCALE, /* exp(-1/180) */
86 };
87
88 /* kernel uses `FSCALE', userland (SHOULD) use kern.fscale */
89 static int fscale __unused = FSCALE;
90 SYSCTL_INT(_kern, OID_AUTO, fscale, CTLFLAG_RD, 0, FSCALE, "");
91
92 static void loadav(void *arg);
93 static void lboltcb(void *arg);
94
95 void
96 sleepinit(void)
97 {
98
99 hogticks = (hz / 10) * 2; /* Default only. */
100 init_sleepqueues();
101 }
102
103 /*
104 * General sleep call. Suspends the current process until a wakeup is
105 * performed on the specified identifier. The process will then be made
106 * runnable with the specified priority. Sleeps at most timo/hz seconds
107 * (0 means no timeout). If pri includes PCATCH flag, signals are checked
108 * before and after sleeping, else signals are not checked. Returns 0 if
109 * awakened, EWOULDBLOCK if the timeout expires. If PCATCH is set and a
110 * signal needs to be delivered, ERESTART is returned if the current system
111 * call should be restarted if possible, and EINTR is returned if the system
112 * call should be interrupted by the signal (return EINTR).
113 *
114 * The mutex argument is exited before the caller is suspended, and
115 * entered before msleep returns. If priority includes the PDROP
116 * flag the mutex is not entered before returning.
117 */
118 int
119 msleep(ident, mtx, priority, wmesg, timo)
120 void *ident;
121 struct mtx *mtx;
122 int priority, timo;
123 const char *wmesg;
124 {
125 struct thread *td;
126 struct proc *p;
127 int catch, rval, flags;
128 WITNESS_SAVE_DECL(mtx);
129
130 td = curthread;
131 p = td->td_proc;
132 #ifdef KTRACE
133 if (KTRPOINT(td, KTR_CSW))
134 ktrcsw(1, 0);
135 #endif
136 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, mtx == NULL ? NULL :
137 &mtx->mtx_object, "Sleeping on \"%s\"", wmesg);
138 KASSERT(timo != 0 || mtx_owned(&Giant) || mtx != NULL,
139 ("sleeping without a mutex"));
140 KASSERT(p != NULL, ("msleep1"));
141 KASSERT(ident != NULL && TD_IS_RUNNING(td), ("msleep"));
142
143 if (cold) {
144 /*
145 * During autoconfiguration, just return;
146 * don't run any other threads or panic below,
147 * in case this is the idle thread and already asleep.
148 * XXX: this used to do "s = splhigh(); splx(safepri);
149 * splx(s);" to give interrupts a chance, but there is
150 * no way to give interrupts a chance now.
151 */
152 if (mtx != NULL && priority & PDROP)
153 mtx_unlock(mtx);
154 return (0);
155 }
156 catch = priority & PCATCH;
157 rval = 0;
158
159 /*
160 * If we are already on a sleep queue, then remove us from that
161 * sleep queue first. We have to do this to handle recursive
162 * sleeps.
163 */
164 if (TD_ON_SLEEPQ(td))
165 sleepq_remove(td, td->td_wchan);
166
167 flags = SLEEPQ_MSLEEP;
168 if (catch)
169 flags |= SLEEPQ_INTERRUPTIBLE;
170
171 sleepq_lock(ident);
172 CTR5(KTR_PROC, "msleep: thread %p (pid %ld, %s) on %s (%p)",
173 (void *)td, (long)p->p_pid, p->p_comm, wmesg, ident);
174
175 DROP_GIANT();
176 if (mtx != NULL) {
177 mtx_assert(mtx, MA_OWNED | MA_NOTRECURSED);
178 WITNESS_SAVE(&mtx->mtx_object, mtx);
179 mtx_unlock(mtx);
180 }
181
182 /*
183 * We put ourselves on the sleep queue and start our timeout
184 * before calling thread_suspend_check, as we could stop there,
185 * and a wakeup or a SIGCONT (or both) could occur while we were
186 * stopped without resuming us. Thus, we must be ready for sleep
187 * when cursig() is called. If the wakeup happens while we're
188 * stopped, then td will no longer be on a sleep queue upon
189 * return from cursig().
190 */
191 sleepq_add(ident, mtx, wmesg, flags);
192 if (timo)
193 sleepq_set_timeout(ident, timo);
194
195 /*
196 * Adjust this thread's priority.
197 */
198 mtx_lock_spin(&sched_lock);
199 sched_prio(td, priority & PRIMASK);
200 mtx_unlock_spin(&sched_lock);
201
202 if (timo && catch)
203 rval = sleepq_timedwait_sig(ident);
204 else if (timo)
205 rval = sleepq_timedwait(ident);
206 else if (catch)
207 rval = sleepq_wait_sig(ident);
208 else {
209 sleepq_wait(ident);
210 rval = 0;
211 }
212 #ifdef KTRACE
213 if (KTRPOINT(td, KTR_CSW))
214 ktrcsw(0, 0);
215 #endif
216 PICKUP_GIANT();
217 if (mtx != NULL && !(priority & PDROP)) {
218 mtx_lock(mtx);
219 WITNESS_RESTORE(&mtx->mtx_object, mtx);
220 }
221 return (rval);
222 }
223
224 /*
225 * Make all threads sleeping on the specified identifier runnable.
226 */
227 void
228 wakeup(ident)
229 register void *ident;
230 {
231
232 sleepq_lock(ident);
233 sleepq_broadcast(ident, SLEEPQ_MSLEEP, -1);
234 }
235
236 /*
237 * Make a thread sleeping on the specified identifier runnable.
238 * May wake more than one thread if a target thread is currently
239 * swapped out.
240 */
241 void
242 wakeup_one(ident)
243 register void *ident;
244 {
245
246 sleepq_lock(ident);
247 sleepq_signal(ident, SLEEPQ_MSLEEP, -1);
248 }
249
250 /*
251 * The machine independent parts of context switching.
252 */
253 void
254 mi_switch(int flags, struct thread *newtd)
255 {
256 struct bintime new_switchtime;
257 struct thread *td;
258 struct proc *p;
259
260 mtx_assert(&sched_lock, MA_OWNED | MA_NOTRECURSED);
261 td = curthread; /* XXX */
262 p = td->td_proc; /* XXX */
263 KASSERT(!TD_ON_RUNQ(td), ("mi_switch: called by old code"));
264 #ifdef INVARIANTS
265 if (!TD_ON_LOCK(td) && !TD_IS_RUNNING(td))
266 mtx_assert(&Giant, MA_NOTOWNED);
267 #endif
268 KASSERT(td->td_critnest == 1 || (td->td_critnest == 2 &&
269 (td->td_owepreempt) && (flags & SW_INVOL) != 0 &&
270 newtd == NULL) || panicstr,
271 ("mi_switch: switch in a critical section"));
272 KASSERT((flags & (SW_INVOL | SW_VOL)) != 0,
273 ("mi_switch: switch must be voluntary or involuntary"));
274 KASSERT(newtd != curthread, ("mi_switch: preempting back to ourself"));
275
276 if (flags & SW_VOL)
277 p->p_stats->p_ru.ru_nvcsw++;
278 else
279 p->p_stats->p_ru.ru_nivcsw++;
280
281 /*
282 * Compute the amount of time during which the current
283 * process was running, and add that to its total so far.
284 */
285 binuptime(&new_switchtime);
286 bintime_add(&p->p_rux.rux_runtime, &new_switchtime);
287 bintime_sub(&p->p_rux.rux_runtime, PCPU_PTR(switchtime));
288
289 td->td_generation++; /* bump preempt-detect counter */
290
291 /*
292 * Don't perform context switches from the debugger.
293 */
294 if (kdb_active) {
295 mtx_unlock_spin(&sched_lock);
296 kdb_backtrace();
297 kdb_reenter();
298 panic("%s: did not reenter debugger", __func__);
299 }
300
301 /*
302 * Check if the process exceeds its cpu resource allocation. If
303 * it reaches the max, arrange to kill the process in ast().
304 */
305 if (p->p_cpulimit != RLIM_INFINITY &&
306 p->p_rux.rux_runtime.sec >= p->p_cpulimit) {
307 p->p_sflag |= PS_XCPU;
308 td->td_flags |= TDF_ASTPENDING;
309 }
310
311 /*
312 * Finish up stats for outgoing thread.
313 */
314 cnt.v_swtch++;
315 PCPU_SET(switchtime, new_switchtime);
316 PCPU_SET(switchticks, ticks);
317 CTR4(KTR_PROC, "mi_switch: old thread %p (kse %p, pid %ld, %s)",
318 (void *)td, td->td_sched, (long)p->p_pid, p->p_comm);
319 if ((flags & SW_VOL) && (td->td_proc->p_flag & P_SA))
320 newtd = thread_switchout(td, flags, newtd);
321 #if (KTR_COMPILE & KTR_SCHED) != 0
322 if (td == PCPU_GET(idlethread))
323 CTR3(KTR_SCHED, "mi_switch: %p(%s) prio %d idle",
324 td, td->td_proc->p_comm, td->td_priority);
325 else if (newtd != NULL)
326 CTR5(KTR_SCHED,
327 "mi_switch: %p(%s) prio %d preempted by %p(%s)",
328 td, td->td_proc->p_comm, td->td_priority, newtd,
329 newtd->td_proc->p_comm);
330 else
331 CTR6(KTR_SCHED,
332 "mi_switch: %p(%s) prio %d inhibit %d wmesg %s lock %s",
333 td, td->td_proc->p_comm, td->td_priority,
334 td->td_inhibitors, td->td_wmesg, td->td_lockname);
335 #endif
336 sched_switch(td, newtd, flags);
337 CTR3(KTR_SCHED, "mi_switch: running %p(%s) prio %d",
338 td, td->td_proc->p_comm, td->td_priority);
339
340 CTR4(KTR_PROC, "mi_switch: new thread %p (kse %p, pid %ld, %s)",
341 (void *)td, td->td_sched, (long)p->p_pid, p->p_comm);
342
343 /*
344 * If the last thread was exiting, finish cleaning it up.
345 */
346 if ((td = PCPU_GET(deadthread))) {
347 PCPU_SET(deadthread, NULL);
348 thread_stash(td);
349 }
350 }
351
352 /*
353 * Change process state to be runnable,
354 * placing it on the run queue if it is in memory,
355 * and awakening the swapper if it isn't in memory.
356 */
357 void
358 setrunnable(struct thread *td)
359 {
360 struct proc *p;
361
362 p = td->td_proc;
363 mtx_assert(&sched_lock, MA_OWNED);
364 switch (p->p_state) {
365 case PRS_ZOMBIE:
366 panic("setrunnable(1)");
367 default:
368 break;
369 }
370 switch (td->td_state) {
371 case TDS_RUNNING:
372 case TDS_RUNQ:
373 return;
374 case TDS_INHIBITED:
375 /*
376 * If we are only inhibited because we are swapped out
377 * then arange to swap in this process. Otherwise just return.
378 */
379 if (td->td_inhibitors != TDI_SWAPPED)
380 return;
381 /* XXX: intentional fall-through ? */
382 case TDS_CAN_RUN:
383 break;
384 default:
385 printf("state is 0x%x", td->td_state);
386 panic("setrunnable(2)");
387 }
388 if ((p->p_sflag & PS_INMEM) == 0) {
389 if ((p->p_sflag & PS_SWAPPINGIN) == 0) {
390 p->p_sflag |= PS_SWAPINREQ;
391 /*
392 * due to a LOR between sched_lock and
393 * the sleepqueue chain locks, use
394 * lower level scheduling functions.
395 */
396 kick_proc0();
397 }
398 } else
399 sched_wakeup(td);
400 }
401
402 /*
403 * Compute a tenex style load average of a quantity on
404 * 1, 5 and 15 minute intervals.
405 * XXXKSE Needs complete rewrite when correct info is available.
406 * Completely Bogus.. only works with 1:1 (but compiles ok now :-)
407 */
408 static void
409 loadav(void *arg)
410 {
411 int i, nrun;
412 struct loadavg *avg;
413
414 nrun = sched_load();
415 avg = &averunnable;
416
417 for (i = 0; i < 3; i++)
418 avg->ldavg[i] = (cexp[i] * avg->ldavg[i] +
419 nrun * FSCALE * (FSCALE - cexp[i])) >> FSHIFT;
420
421 /*
422 * Schedule the next update to occur after 5 seconds, but add a
423 * random variation to avoid synchronisation with processes that
424 * run at regular intervals.
425 */
426 callout_reset(&loadav_callout, hz * 4 + (int)(random() % (hz * 2 + 1)),
427 loadav, NULL);
428 }
429
430 static void
431 lboltcb(void *arg)
432 {
433 wakeup(&lbolt);
434 callout_reset(&lbolt_callout, hz, lboltcb, NULL);
435 }
436
437 /* ARGSUSED */
438 static void
439 synch_setup(dummy)
440 void *dummy;
441 {
442 callout_init(&loadav_callout, CALLOUT_MPSAFE);
443 callout_init(&lbolt_callout, CALLOUT_MPSAFE);
444
445 /* Kick off timeout driven events by calling first time. */
446 loadav(NULL);
447 lboltcb(NULL);
448 }
449
450 /*
451 * General purpose yield system call
452 */
453 int
454 yield(struct thread *td, struct yield_args *uap)
455 {
456 struct ksegrp *kg;
457
458 kg = td->td_ksegrp;
459 mtx_assert(&Giant, MA_NOTOWNED);
460 mtx_lock_spin(&sched_lock);
461 sched_prio(td, PRI_MAX_TIMESHARE);
462 mi_switch(SW_VOL, NULL);
463 mtx_unlock_spin(&sched_lock);
464 td->td_retval[0] = 0;
465 return (0);
466 }
Cache object: ba3b986083f7f46cfa0a25c605da48aa
|