kern_timeout.c revision 1.12 1 /* $NetBSD: kern_timeout.c,v 1.12 2003/10/27 16:52:01 thorpej Exp $ */
2
3 /*-
4 * Copyright (c) 2003 The NetBSD Foundation, Inc.
5 * All rights reserved.
6 *
7 * This code is derived from software contributed to The NetBSD Foundation
8 * by Jason R. Thorpe.
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 * 3. All advertising materials mentioning features or use of this software
19 * must display the following acknowledgement:
20 * This product includes software developed by the NetBSD
21 * Foundation, Inc. and its contributors.
22 * 4. Neither the name of The NetBSD Foundation nor the names of its
23 * contributors may be used to endorse or promote products derived
24 * from this software without specific prior written permission.
25 *
26 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
27 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
28 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
29 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
30 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
31 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
32 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
33 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
34 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
35 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
36 * POSSIBILITY OF SUCH DAMAGE.
37 */
38
39 /*
40 * Copyright (c) 2001 Thomas Nordin <nordin (at) openbsd.org>
41 * Copyright (c) 2000-2001 Artur Grabowski <art (at) openbsd.org>
42 * All rights reserved.
43 *
44 * Redistribution and use in source and binary forms, with or without
45 * modification, are permitted provided that the following conditions
46 * are met:
47 *
48 * 1. Redistributions of source code must retain the above copyright
49 * notice, this list of conditions and the following disclaimer.
50 * 2. Redistributions in binary form must reproduce the above copyright
51 * notice, this list of conditions and the following disclaimer in the
52 * documentation and/or other materials provided with the distribution.
53 * 3. The name of the author may not be used to endorse or promote products
54 * derived from this software without specific prior written permission.
55 *
56 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES,
57 * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY
58 * AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL
59 * THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
60 * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
61 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;
62 * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
63 * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
64 * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
65 * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
66 */
67
68 #include <sys/cdefs.h>
69 __KERNEL_RCSID(0, "$NetBSD: kern_timeout.c,v 1.12 2003/10/27 16:52:01 thorpej Exp $");
70
71 /*
72 * Adapted from OpenBSD: kern_timeout.c,v 1.15 2002/12/08 04:21:07 art Exp,
73 * modified to match NetBSD's pre-existing callout API.
74 */
75
76 #include <sys/param.h>
77 #include <sys/systm.h>
78 #include <sys/kernel.h>
79 #include <sys/lock.h>
80 #include <sys/callout.h>
81
82 #ifdef DDB
83 #include <machine/db_machdep.h>
84 #include <ddb/db_interface.h>
85 #include <ddb/db_access.h>
86 #include <ddb/db_sym.h>
87 #include <ddb/db_output.h>
88 #endif
89
90 /*
91 * Timeouts are kept in a hierarchical timing wheel. The c_time is the value
92 * of the global variable "hardclock_ticks" when the timeout should be called.
93 * There are four levels with 256 buckets each. See 'Scheme 7' in
94 * "Hashed and Hierarchical Timing Wheels: Efficient Data Structures for
95 * Implementing a Timer Facility" by George Varghese and Tony Lauck.
96 */
97 #define BUCKETS 1024
98 #define WHEELSIZE 256
99 #define WHEELMASK 255
100 #define WHEELBITS 8
101
102 static struct callout_circq timeout_wheel[BUCKETS]; /* Queues of timeouts */
103 static struct callout_circq timeout_todo; /* Worklist */
104
105 #define MASKWHEEL(wheel, time) (((time) >> ((wheel)*WHEELBITS)) & WHEELMASK)
106
107 #define BUCKET(rel, abs) \
108 (((rel) <= (1 << (2*WHEELBITS))) \
109 ? ((rel) <= (1 << WHEELBITS)) \
110 ? &timeout_wheel[MASKWHEEL(0, (abs))] \
111 : &timeout_wheel[MASKWHEEL(1, (abs)) + WHEELSIZE] \
112 : ((rel) <= (1 << (3*WHEELBITS))) \
113 ? &timeout_wheel[MASKWHEEL(2, (abs)) + 2*WHEELSIZE] \
114 : &timeout_wheel[MASKWHEEL(3, (abs)) + 3*WHEELSIZE])
115
116 #define MOVEBUCKET(wheel, time) \
117 CIRCQ_APPEND(&timeout_todo, \
118 &timeout_wheel[MASKWHEEL((wheel), (time)) + (wheel)*WHEELSIZE])
119
120 /*
121 * All wheels are locked with the same lock (which must also block out all
122 * interrupts).
123 */
124 static struct simplelock callout_slock;
125
126 #define CALLOUT_LOCK(s) \
127 do { \
128 s = splsched(); \
129 simple_lock(&callout_slock); \
130 } while (/*CONSTCOND*/0)
131
132 #define CALLOUT_UNLOCK(s) \
133 do { \
134 simple_unlock(&callout_slock); \
135 splx((s)); \
136 } while (/*CONSTCOND*/0)
137
138 /*
139 * Circular queue definitions.
140 */
141
142 #define CIRCQ_INIT(list) \
143 do { \
144 (list)->cq_next_l = (list); \
145 (list)->cq_prev_l = (list); \
146 } while (/*CONSTCOND*/0)
147
148 #define CIRCQ_INSERT(elem, list) \
149 do { \
150 (elem)->cq_prev_e = (list)->cq_prev_e; \
151 (elem)->cq_next_l = (list); \
152 (list)->cq_prev_l->cq_next_l = (elem); \
153 (list)->cq_prev_l = (elem); \
154 } while (/*CONSTCOND*/0)
155
156 #define CIRCQ_APPEND(fst, snd) \
157 do { \
158 if (!CIRCQ_EMPTY(snd)) { \
159 (fst)->cq_prev_l->cq_next_l = (snd)->cq_next_l; \
160 (snd)->cq_next_l->cq_prev_l = (fst)->cq_prev_l; \
161 (snd)->cq_prev_l->cq_next_l = (fst); \
162 (fst)->cq_prev_l = (snd)->cq_prev_l; \
163 CIRCQ_INIT(snd); \
164 } \
165 } while (/*CONSTCOND*/0)
166
167 #define CIRCQ_REMOVE(elem) \
168 do { \
169 (elem)->cq_next_l->cq_prev_e = (elem)->cq_prev_e; \
170 (elem)->cq_prev_l->cq_next_e = (elem)->cq_next_e; \
171 } while (/*CONSTCOND*/0)
172
173 #define CIRCQ_FIRST(list) ((list)->cq_next_e)
174 #define CIRCQ_NEXT(elem) ((elem)->cq_next_e)
175 #define CIRCQ_LAST(elem,list) ((elem)->cq_next_l == (list))
176 #define CIRCQ_EMPTY(list) ((list)->cq_next_l == (list))
177
178 /*
179 * Some of the "math" in here is a bit tricky.
180 *
181 * We have to beware of wrapping ints.
182 * We use the fact that any element added to the queue must be added with a
183 * positive time. That means that any element `to' on the queue cannot be
184 * scheduled to timeout further in time than INT_MAX, but c->c_time can
185 * be positive or negative so comparing it with anything is dangerous.
186 * The only way we can use the c->c_time value in any predictable way
187 * is when we calculate how far in the future `to' will timeout -
188 * "c->c_time - hardclock_ticks". The result will always be positive for
189 * future timeouts and 0 or negative for due timeouts.
190 */
191
192 #ifdef CALLOUT_EVENT_COUNTERS
193 static struct evcnt callout_ev_late;
194 #endif
195
196 /*
197 * callout_startup:
198 *
199 * Initialize the callout facility, called at system startup time.
200 */
201 void
202 callout_startup(void)
203 {
204 int b;
205
206 CIRCQ_INIT(&timeout_todo);
207 for (b = 0; b < BUCKETS; b++)
208 CIRCQ_INIT(&timeout_wheel[b]);
209 simple_lock_init(&callout_slock);
210
211 #ifdef CALLOUT_EVENT_COUNTERS
212 evcnt_attach_dynamic(&callout_ev_late, EVCNT_TYPE_MISC,
213 NULL, "callout", "late");
214 #endif
215 }
216
217 /*
218 * callout_init:
219 *
220 * Initialize a callout structure.
221 */
222 void
223 callout_init(struct callout *c)
224 {
225
226 memset(c, 0, sizeof(*c));
227 }
228
229 /*
230 * callout_setfunc:
231 *
232 * Initialize a callout structure and set the function and
233 * argument.
234 *
235 * NOTE: THE CALLOUT STRUCTURE MUST ALREADY BE INITIALIZED!
236 */
237 void
238 callout_setfunc(struct callout *c, void (*func)(void *), void *arg)
239 {
240
241 c->c_func = func;
242 c->c_arg = arg;
243 }
244
245 /*
246 * callout_reset:
247 *
248 * Reset a callout structure with a new function and argument, and
249 * schedule it to run.
250 */
251 void
252 callout_reset(struct callout *c, int to_ticks, void (*func)(void *), void *arg)
253 {
254 int s, old_time;
255
256 KASSERT(to_ticks >= 0);
257
258 CALLOUT_LOCK(s);
259
260 /* Initialize the time here, it won't change. */
261 old_time = c->c_time;
262 c->c_time = to_ticks + hardclock_ticks;
263 c->c_flags &= ~(CALLOUT_FIRED|CALLOUT_INVOKING);
264
265 c->c_func = func;
266 c->c_arg = arg;
267
268 /*
269 * If this timeout is already scheduled and now is moved
270 * earlier, reschedule it now. Otherwise leave it in place
271 * and let it be rescheduled later.
272 */
273 if (callout_pending(c)) {
274 if (c->c_time - old_time < 0) {
275 CIRCQ_REMOVE(&c->c_list);
276 CIRCQ_INSERT(&c->c_list, &timeout_todo);
277 }
278 } else {
279 c->c_flags |= CALLOUT_PENDING;
280 CIRCQ_INSERT(&c->c_list, &timeout_todo);
281 }
282
283 CALLOUT_UNLOCK(s);
284 }
285
286 /*
287 * callout_schedule:
288 *
289 * Schedule a callout to run. The function and argument must
290 * already be set in the callout structure.
291 */
292 void
293 callout_schedule(struct callout *c, int to_ticks)
294 {
295 int s, old_time;
296
297 KASSERT(to_ticks >= 0);
298
299 CALLOUT_LOCK(s);
300
301 /* Initialize the time here, it won't change. */
302 old_time = c->c_time;
303 c->c_time = to_ticks + hardclock_ticks;
304 c->c_flags &= ~(CALLOUT_FIRED|CALLOUT_INVOKING);
305
306 /*
307 * If this timeout is already scheduled and now is moved
308 * earlier, reschedule it now. Otherwise leave it in place
309 * and let it be rescheduled later.
310 */
311 if (callout_pending(c)) {
312 if (c->c_time - old_time < 0) {
313 CIRCQ_REMOVE(&c->c_list);
314 CIRCQ_INSERT(&c->c_list, &timeout_todo);
315 }
316 } else {
317 c->c_flags |= CALLOUT_PENDING;
318 CIRCQ_INSERT(&c->c_list, &timeout_todo);
319 }
320
321 CALLOUT_UNLOCK(s);
322 }
323
324 /*
325 * callout_stop:
326 *
327 * Cancel a pending callout.
328 */
329 void
330 callout_stop(struct callout *c)
331 {
332 int s;
333
334 CALLOUT_LOCK(s);
335
336 if (callout_pending(c))
337 CIRCQ_REMOVE(&c->c_list);
338
339 c->c_flags &= ~(CALLOUT_PENDING|CALLOUT_FIRED);
340
341 CALLOUT_UNLOCK(s);
342 }
343
344 /*
345 * This is called from hardclock() once every tick.
346 * We return !0 if we need to schedule a softclock.
347 */
348 int
349 callout_hardclock(void)
350 {
351 int s;
352 int needsoftclock;
353
354 CALLOUT_LOCK(s);
355
356 MOVEBUCKET(0, hardclock_ticks);
357 if (MASKWHEEL(0, hardclock_ticks) == 0) {
358 MOVEBUCKET(1, hardclock_ticks);
359 if (MASKWHEEL(1, hardclock_ticks) == 0) {
360 MOVEBUCKET(2, hardclock_ticks);
361 if (MASKWHEEL(2, hardclock_ticks) == 0)
362 MOVEBUCKET(3, hardclock_ticks);
363 }
364 }
365
366 needsoftclock = !CIRCQ_EMPTY(&timeout_todo);
367 CALLOUT_UNLOCK(s);
368
369 return needsoftclock;
370 }
371
372 /* ARGSUSED */
373 void
374 softclock(void *v)
375 {
376 struct callout *c;
377 void (*func)(void *);
378 void *arg;
379 int s;
380
381 CALLOUT_LOCK(s);
382
383 while (!CIRCQ_EMPTY(&timeout_todo)) {
384 c = CIRCQ_FIRST(&timeout_todo);
385 CIRCQ_REMOVE(&c->c_list);
386
387 /* If due run it, otherwise insert it into the right bucket. */
388 if (c->c_time - hardclock_ticks > 0) {
389 CIRCQ_INSERT(&c->c_list,
390 BUCKET((c->c_time - hardclock_ticks), c->c_time));
391 } else {
392 #ifdef CALLOUT_EVENT_COUNTERS
393 if (c->c_time - hardclock_ticks < 0)
394 callout_ev_late.ev_count++;
395 #endif
396 c->c_flags = (c->c_flags & ~CALLOUT_PENDING) |
397 (CALLOUT_FIRED|CALLOUT_INVOKING);
398
399 func = c->c_func;
400 arg = c->c_arg;
401
402 CALLOUT_UNLOCK(s);
403 (*func)(arg);
404 CALLOUT_LOCK(s);
405 }
406 }
407
408 CALLOUT_UNLOCK(s);
409 }
410
411 #ifdef DDB
412 static void
413 db_show_callout_bucket(struct callout_circq *bucket)
414 {
415 struct callout *c;
416 db_expr_t offset;
417 char *name;
418
419 if (CIRCQ_EMPTY(bucket))
420 return;
421
422 for (c = CIRCQ_FIRST(bucket); /*nothing*/; c = CIRCQ_NEXT(&c->c_list)) {
423 db_find_sym_and_offset((db_addr_t)(intptr_t)c->c_func, &name,
424 &offset);
425 name = name ? name : "?";
426 #ifdef _LP64
427 #define POINTER_WIDTH "%16lx"
428 #else
429 #define POINTER_WIDTH "%8lx"
430 #endif
431 db_printf("%9d %2d/%-4d " POINTER_WIDTH " %s\n",
432 c->c_time - hardclock_ticks,
433 (int)((bucket - timeout_wheel) / WHEELSIZE),
434 (int)(bucket - timeout_wheel), (u_long) c->c_arg, name);
435
436 if (CIRCQ_LAST(&c->c_list, bucket))
437 break;
438 }
439 }
440
441 void
442 db_show_callout(db_expr_t addr, int haddr, db_expr_t count, char *modif)
443 {
444 int b;
445
446 db_printf("hardclock_ticks now: %d\n", hardclock_ticks);
447 #ifdef _LP64
448 db_printf(" ticks wheel arg func\n");
449 #else
450 db_printf(" ticks wheel arg func\n");
451 #endif
452
453 /*
454 * Don't lock the callwheel; all the other CPUs are paused
455 * anyhow, and we might be called in a circumstance where
456 * some other CPU was paused while holding the lock.
457 */
458
459 db_show_callout_bucket(&timeout_todo);
460 for (b = 0; b < BUCKETS; b++)
461 db_show_callout_bucket(&timeout_wheel[b]);
462 }
463 #endif /* DDB */
464