4 * Copyright (c) 2003-2008 Fabrice Bellard
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to deal
8 * in the Software without restriction, including without limitation the rights
9 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10 * copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
25 #include "qemu/osdep.h"
26 #include "qemu/main-loop.h"
27 #include "qemu/timer.h"
28 #include "qemu/lockable.h"
29 #include "sysemu/replay.h"
30 #include "sysemu/cpus.h"
40 #ifdef CONFIG_PRCTL_PR_SET_TIMERSLACK
41 #include <sys/prctl.h>
44 /***********************************************************/
47 typedef struct QEMUClock
{
48 /* We rely on BQL to protect the timerlists */
49 QLIST_HEAD(, QEMUTimerList
) timerlists
;
55 QEMUTimerListGroup main_loop_tlg
;
56 static QEMUClock qemu_clocks
[QEMU_CLOCK_MAX
];
58 /* A QEMUTimerList is a list of timers attached to a clock. More
59 * than one QEMUTimerList can be attached to each clock, for instance
60 * used by different AioContexts / threads. Each clock also has
61 * a list of the QEMUTimerLists associated with it, in order that
62 * reenabling the clock can call all the notifiers.
65 struct QEMUTimerList
{
67 QemuMutex active_timers_lock
;
68 QEMUTimer
*active_timers
;
69 QLIST_ENTRY(QEMUTimerList
) list
;
70 QEMUTimerListNotifyCB
*notify_cb
;
73 /* lightweight method to mark the end of timerlist's running */
74 QemuEvent timers_done_ev
;
79 * @type: type of clock
81 * Translate a clock type into a pointer to QEMUClock object.
83 * Returns: a pointer to the QEMUClock object
85 static inline QEMUClock
*qemu_clock_ptr(QEMUClockType type
)
87 return &qemu_clocks
[type
];
90 static bool timer_expired_ns(QEMUTimer
*timer_head
, int64_t current_time
)
92 return timer_head
&& (timer_head
->expire_time
<= current_time
);
95 QEMUTimerList
*timerlist_new(QEMUClockType type
,
96 QEMUTimerListNotifyCB
*cb
,
99 QEMUTimerList
*timer_list
;
100 QEMUClock
*clock
= qemu_clock_ptr(type
);
102 timer_list
= g_malloc0(sizeof(QEMUTimerList
));
103 qemu_event_init(&timer_list
->timers_done_ev
, true);
104 timer_list
->clock
= clock
;
105 timer_list
->notify_cb
= cb
;
106 timer_list
->notify_opaque
= opaque
;
107 qemu_mutex_init(&timer_list
->active_timers_lock
);
108 QLIST_INSERT_HEAD(&clock
->timerlists
, timer_list
, list
);
112 void timerlist_free(QEMUTimerList
*timer_list
)
114 assert(!timerlist_has_timers(timer_list
));
115 if (timer_list
->clock
) {
116 QLIST_REMOVE(timer_list
, list
);
118 qemu_mutex_destroy(&timer_list
->active_timers_lock
);
122 static void qemu_clock_init(QEMUClockType type
, QEMUTimerListNotifyCB
*notify_cb
)
124 QEMUClock
*clock
= qemu_clock_ptr(type
);
126 /* Assert that the clock of type TYPE has not been initialized yet. */
127 assert(main_loop_tlg
.tl
[type
] == NULL
);
130 clock
->enabled
= (type
== QEMU_CLOCK_VIRTUAL
? false : true);
131 QLIST_INIT(&clock
->timerlists
);
132 main_loop_tlg
.tl
[type
] = timerlist_new(type
, notify_cb
, NULL
);
135 bool qemu_clock_use_for_deadline(QEMUClockType type
)
137 return !(use_icount
&& (type
== QEMU_CLOCK_VIRTUAL
));
140 void qemu_clock_notify(QEMUClockType type
)
142 QEMUTimerList
*timer_list
;
143 QEMUClock
*clock
= qemu_clock_ptr(type
);
144 QLIST_FOREACH(timer_list
, &clock
->timerlists
, list
) {
145 timerlist_notify(timer_list
);
149 /* Disabling the clock will wait for related timerlists to stop
150 * executing qemu_run_timers. Thus, this functions should not
151 * be used from the callback of a timer that is based on @clock.
152 * Doing so would cause a deadlock.
154 * Caller should hold BQL.
156 void qemu_clock_enable(QEMUClockType type
, bool enabled
)
158 QEMUClock
*clock
= qemu_clock_ptr(type
);
160 bool old
= clock
->enabled
;
161 clock
->enabled
= enabled
;
162 if (enabled
&& !old
) {
163 qemu_clock_notify(type
);
164 } else if (!enabled
&& old
) {
165 QLIST_FOREACH(tl
, &clock
->timerlists
, list
) {
166 qemu_event_wait(&tl
->timers_done_ev
);
171 bool timerlist_has_timers(QEMUTimerList
*timer_list
)
173 return !!atomic_read(&timer_list
->active_timers
);
176 bool qemu_clock_has_timers(QEMUClockType type
)
178 return timerlist_has_timers(
179 main_loop_tlg
.tl
[type
]);
182 bool timerlist_expired(QEMUTimerList
*timer_list
)
186 if (!atomic_read(&timer_list
->active_timers
)) {
190 WITH_QEMU_LOCK_GUARD(&timer_list
->active_timers_lock
) {
191 if (!timer_list
->active_timers
) {
194 expire_time
= timer_list
->active_timers
->expire_time
;
197 return expire_time
<= qemu_clock_get_ns(timer_list
->clock
->type
);
200 bool qemu_clock_expired(QEMUClockType type
)
202 return timerlist_expired(
203 main_loop_tlg
.tl
[type
]);
207 * As above, but return -1 for no deadline, and do not cap to 2^32
208 * as we know the result is always positive.
211 int64_t timerlist_deadline_ns(QEMUTimerList
*timer_list
)
216 if (!atomic_read(&timer_list
->active_timers
)) {
220 if (!timer_list
->clock
->enabled
) {
224 /* The active timers list may be modified before the caller uses our return
225 * value but ->notify_cb() is called when the deadline changes. Therefore
226 * the caller should notice the change and there is no race condition.
228 WITH_QEMU_LOCK_GUARD(&timer_list
->active_timers_lock
) {
229 if (!timer_list
->active_timers
) {
232 expire_time
= timer_list
->active_timers
->expire_time
;
235 delta
= expire_time
- qemu_clock_get_ns(timer_list
->clock
->type
);
244 /* Calculate the soonest deadline across all timerlists attached
245 * to the clock. This is used for the icount timeout so we
246 * ignore whether or not the clock should be used in deadline
249 int64_t qemu_clock_deadline_ns_all(QEMUClockType type
, int attr_mask
)
251 int64_t deadline
= -1;
255 QEMUTimerList
*timer_list
;
256 QEMUClock
*clock
= qemu_clock_ptr(type
);
258 if (!clock
->enabled
) {
262 QLIST_FOREACH(timer_list
, &clock
->timerlists
, list
) {
263 qemu_mutex_lock(&timer_list
->active_timers_lock
);
264 ts
= timer_list
->active_timers
;
265 /* Skip all external timers */
266 while (ts
&& (ts
->attributes
& ~attr_mask
)) {
270 qemu_mutex_unlock(&timer_list
->active_timers_lock
);
273 expire_time
= ts
->expire_time
;
274 qemu_mutex_unlock(&timer_list
->active_timers_lock
);
276 delta
= expire_time
- qemu_clock_get_ns(type
);
280 deadline
= qemu_soonest_timeout(deadline
, delta
);
285 QEMUClockType
timerlist_get_clock(QEMUTimerList
*timer_list
)
287 return timer_list
->clock
->type
;
290 QEMUTimerList
*qemu_clock_get_main_loop_timerlist(QEMUClockType type
)
292 return main_loop_tlg
.tl
[type
];
295 void timerlist_notify(QEMUTimerList
*timer_list
)
297 if (timer_list
->notify_cb
) {
298 timer_list
->notify_cb(timer_list
->notify_opaque
, timer_list
->clock
->type
);
304 /* Transition function to convert a nanosecond timeout to ms
305 * This is used where a system does not support ppoll
307 int qemu_timeout_ns_to_ms(int64_t ns
)
318 /* Always round up, because it's better to wait too long than to wait too
319 * little and effectively busy-wait
321 ms
= DIV_ROUND_UP(ns
, SCALE_MS
);
323 /* To avoid overflow problems, limit this to 2^31, i.e. approx 25 days */
324 return MIN(ms
, INT32_MAX
);
328 /* qemu implementation of g_poll which uses a nanosecond timeout but is
329 * otherwise identical to g_poll
331 int qemu_poll_ns(GPollFD
*fds
, guint nfds
, int64_t timeout
)
335 return ppoll((struct pollfd
*)fds
, nfds
, NULL
, NULL
);
338 int64_t tvsec
= timeout
/ 1000000000LL;
339 /* Avoid possibly overflowing and specifying a negative number of
340 * seconds, which would turn a very long timeout into a busy-wait.
342 if (tvsec
> (int64_t)INT32_MAX
) {
346 ts
.tv_nsec
= timeout
% 1000000000LL;
347 return ppoll((struct pollfd
*)fds
, nfds
, &ts
, NULL
);
350 return g_poll(fds
, nfds
, qemu_timeout_ns_to_ms(timeout
));
355 void timer_init_full(QEMUTimer
*ts
,
356 QEMUTimerListGroup
*timer_list_group
, QEMUClockType type
,
357 int scale
, int attributes
,
358 QEMUTimerCB
*cb
, void *opaque
)
360 if (!timer_list_group
) {
361 timer_list_group
= &main_loop_tlg
;
363 ts
->timer_list
= timer_list_group
->tl
[type
];
367 ts
->attributes
= attributes
;
368 ts
->expire_time
= -1;
371 void timer_deinit(QEMUTimer
*ts
)
373 assert(ts
->expire_time
== -1);
374 ts
->timer_list
= NULL
;
377 static void timer_del_locked(QEMUTimerList
*timer_list
, QEMUTimer
*ts
)
381 ts
->expire_time
= -1;
382 pt
= &timer_list
->active_timers
;
388 atomic_set(pt
, t
->next
);
395 static bool timer_mod_ns_locked(QEMUTimerList
*timer_list
,
396 QEMUTimer
*ts
, int64_t expire_time
)
400 /* add the timer in the sorted list */
401 pt
= &timer_list
->active_timers
;
404 if (!timer_expired_ns(t
, expire_time
)) {
409 ts
->expire_time
= MAX(expire_time
, 0);
413 return pt
== &timer_list
->active_timers
;
416 static void timerlist_rearm(QEMUTimerList
*timer_list
)
418 /* Interrupt execution to force deadline recalculation. */
419 if (timer_list
->clock
->type
== QEMU_CLOCK_VIRTUAL
) {
420 qemu_start_warp_timer();
422 timerlist_notify(timer_list
);
425 /* stop a timer, but do not dealloc it */
426 void timer_del(QEMUTimer
*ts
)
428 QEMUTimerList
*timer_list
= ts
->timer_list
;
431 qemu_mutex_lock(&timer_list
->active_timers_lock
);
432 timer_del_locked(timer_list
, ts
);
433 qemu_mutex_unlock(&timer_list
->active_timers_lock
);
437 /* modify the current timer so that it will be fired when current_time
438 >= expire_time. The corresponding callback will be called. */
439 void timer_mod_ns(QEMUTimer
*ts
, int64_t expire_time
)
441 QEMUTimerList
*timer_list
= ts
->timer_list
;
444 qemu_mutex_lock(&timer_list
->active_timers_lock
);
445 timer_del_locked(timer_list
, ts
);
446 rearm
= timer_mod_ns_locked(timer_list
, ts
, expire_time
);
447 qemu_mutex_unlock(&timer_list
->active_timers_lock
);
450 timerlist_rearm(timer_list
);
454 /* modify the current timer so that it will be fired when current_time
455 >= expire_time or the current deadline, whichever comes earlier.
456 The corresponding callback will be called. */
457 void timer_mod_anticipate_ns(QEMUTimer
*ts
, int64_t expire_time
)
459 QEMUTimerList
*timer_list
= ts
->timer_list
;
462 WITH_QEMU_LOCK_GUARD(&timer_list
->active_timers_lock
) {
463 if (ts
->expire_time
== -1 || ts
->expire_time
> expire_time
) {
464 if (ts
->expire_time
!= -1) {
465 timer_del_locked(timer_list
, ts
);
467 rearm
= timer_mod_ns_locked(timer_list
, ts
, expire_time
);
473 timerlist_rearm(timer_list
);
477 void timer_mod(QEMUTimer
*ts
, int64_t expire_time
)
479 timer_mod_ns(ts
, expire_time
* ts
->scale
);
482 void timer_mod_anticipate(QEMUTimer
*ts
, int64_t expire_time
)
484 timer_mod_anticipate_ns(ts
, expire_time
* ts
->scale
);
487 bool timer_pending(QEMUTimer
*ts
)
489 return ts
->expire_time
>= 0;
492 bool timer_expired(QEMUTimer
*timer_head
, int64_t current_time
)
494 return timer_expired_ns(timer_head
, current_time
* timer_head
->scale
);
497 bool timerlist_run_timers(QEMUTimerList
*timer_list
)
500 int64_t current_time
;
501 bool progress
= false;
504 bool need_replay_checkpoint
= false;
506 if (!atomic_read(&timer_list
->active_timers
)) {
510 qemu_event_reset(&timer_list
->timers_done_ev
);
511 if (!timer_list
->clock
->enabled
) {
515 switch (timer_list
->clock
->type
) {
516 case QEMU_CLOCK_REALTIME
:
519 case QEMU_CLOCK_VIRTUAL
:
520 if (replay_mode
!= REPLAY_MODE_NONE
) {
521 /* Checkpoint for virtual clock is redundant in cases where
522 * it's being triggered with only non-EXTERNAL timers, because
523 * these timers don't change guest state directly.
524 * Since it has conditional dependence on specific timers, it is
525 * subject to race conditions and requires special handling.
528 need_replay_checkpoint
= true;
531 case QEMU_CLOCK_HOST
:
532 if (!replay_checkpoint(CHECKPOINT_CLOCK_HOST
)) {
536 case QEMU_CLOCK_VIRTUAL_RT
:
537 if (!replay_checkpoint(CHECKPOINT_CLOCK_VIRTUAL_RT
)) {
544 * Extract expired timers from active timers list and and process them.
546 * In rr mode we need "filtered" checkpointing for virtual clock. The
547 * checkpoint must be recorded/replayed before processing any non-EXTERNAL timer,
548 * and that must only be done once since the clock value stays the same. Because
549 * non-EXTERNAL timers may appear in the timers list while it being processed,
550 * the checkpoint can be issued at a time until no timers are left and we are
553 current_time
= qemu_clock_get_ns(timer_list
->clock
->type
);
554 qemu_mutex_lock(&timer_list
->active_timers_lock
);
555 while ((ts
= timer_list
->active_timers
)) {
556 if (!timer_expired_ns(ts
, current_time
)) {
557 /* No expired timers left. The checkpoint can be skipped
558 * if no timers fired or they were all external.
562 if (need_replay_checkpoint
563 && !(ts
->attributes
& QEMU_TIMER_ATTR_EXTERNAL
)) {
564 /* once we got here, checkpoint clock only once */
565 need_replay_checkpoint
= false;
566 qemu_mutex_unlock(&timer_list
->active_timers_lock
);
567 if (!replay_checkpoint(CHECKPOINT_CLOCK_VIRTUAL
)) {
570 qemu_mutex_lock(&timer_list
->active_timers_lock
);
571 /* The lock was released; start over again in case the list was
577 /* remove timer from the list before calling the callback */
578 timer_list
->active_timers
= ts
->next
;
580 ts
->expire_time
= -1;
584 /* run the callback (the timer list can be modified) */
585 qemu_mutex_unlock(&timer_list
->active_timers_lock
);
587 qemu_mutex_lock(&timer_list
->active_timers_lock
);
591 qemu_mutex_unlock(&timer_list
->active_timers_lock
);
594 qemu_event_set(&timer_list
->timers_done_ev
);
598 bool qemu_clock_run_timers(QEMUClockType type
)
600 return timerlist_run_timers(main_loop_tlg
.tl
[type
]);
603 void timerlistgroup_init(QEMUTimerListGroup
*tlg
,
604 QEMUTimerListNotifyCB
*cb
, void *opaque
)
607 for (type
= 0; type
< QEMU_CLOCK_MAX
; type
++) {
608 tlg
->tl
[type
] = timerlist_new(type
, cb
, opaque
);
612 void timerlistgroup_deinit(QEMUTimerListGroup
*tlg
)
615 for (type
= 0; type
< QEMU_CLOCK_MAX
; type
++) {
616 timerlist_free(tlg
->tl
[type
]);
620 bool timerlistgroup_run_timers(QEMUTimerListGroup
*tlg
)
623 bool progress
= false;
624 for (type
= 0; type
< QEMU_CLOCK_MAX
; type
++) {
625 progress
|= timerlist_run_timers(tlg
->tl
[type
]);
630 int64_t timerlistgroup_deadline_ns(QEMUTimerListGroup
*tlg
)
632 int64_t deadline
= -1;
634 for (type
= 0; type
< QEMU_CLOCK_MAX
; type
++) {
635 if (qemu_clock_use_for_deadline(type
)) {
636 deadline
= qemu_soonest_timeout(deadline
,
637 timerlist_deadline_ns(tlg
->tl
[type
]));
643 int64_t qemu_clock_get_ns(QEMUClockType type
)
646 case QEMU_CLOCK_REALTIME
:
649 case QEMU_CLOCK_VIRTUAL
:
651 return cpu_get_icount();
653 return cpu_get_clock();
655 case QEMU_CLOCK_HOST
:
656 return REPLAY_CLOCK(REPLAY_CLOCK_HOST
, get_clock_realtime());
657 case QEMU_CLOCK_VIRTUAL_RT
:
658 return REPLAY_CLOCK(REPLAY_CLOCK_VIRTUAL_RT
, cpu_get_clock());
662 void init_clocks(QEMUTimerListNotifyCB
*notify_cb
)
665 for (type
= 0; type
< QEMU_CLOCK_MAX
; type
++) {
666 qemu_clock_init(type
, notify_cb
);
669 #ifdef CONFIG_PRCTL_PR_SET_TIMERSLACK
670 prctl(PR_SET_TIMERSLACK
, 1, 0, 0, 0);
674 uint64_t timer_expire_time_ns(QEMUTimer
*ts
)
676 return timer_pending(ts
) ? ts
->expire_time
: -1;
679 bool qemu_clock_run_all_timers(void)
681 bool progress
= false;
684 for (type
= 0; type
< QEMU_CLOCK_MAX
; type
++) {
685 if (qemu_clock_use_for_deadline(type
)) {
686 progress
|= qemu_clock_run_timers(type
);