1 // SPDX-License-Identifier: GPL-2.0-only
3 * Generic entry points for the idle threads and
4 * implementation of the idle task scheduling class.
6 * (NOTE: these are not related to SCHED_IDLE batch scheduled
7 * tasks which are handled in sched/fair.c )
11 #include <trace/events/power.h>
13 /* Linker adds these: start and end of __cpuidle functions */
14 extern char __cpuidle_text_start
[], __cpuidle_text_end
[];
17 * sched_idle_set_state - Record idle state for the current CPU.
18 * @idle_state: State to record.
20 void sched_idle_set_state(struct cpuidle_state
*idle_state
)
22 idle_set_state(this_rq(), idle_state
);
25 static int __read_mostly cpu_idle_force_poll
;
27 void cpu_idle_poll_ctrl(bool enable
)
30 cpu_idle_force_poll
++;
32 cpu_idle_force_poll
--;
33 WARN_ON_ONCE(cpu_idle_force_poll
< 0);
37 #ifdef CONFIG_GENERIC_IDLE_POLL_SETUP
38 static int __init
cpu_idle_poll_setup(char *__unused
)
40 cpu_idle_force_poll
= 1;
44 __setup("nohlt", cpu_idle_poll_setup
);
46 static int __init
cpu_idle_nopoll_setup(char *__unused
)
48 cpu_idle_force_poll
= 0;
52 __setup("hlt", cpu_idle_nopoll_setup
);
55 static noinline
int __cpuidle
cpu_idle_poll(void)
58 trace_cpu_idle_rcuidle(0, smp_processor_id());
60 stop_critical_timings();
62 while (!tif_need_resched() &&
63 (cpu_idle_force_poll
|| tick_check_broadcast_expired()))
65 start_critical_timings();
66 trace_cpu_idle_rcuidle(PWR_EVENT_EXIT
, smp_processor_id());
72 /* Weak implementations for optional arch specific functions */
73 void __weak
arch_cpu_idle_prepare(void) { }
74 void __weak
arch_cpu_idle_enter(void) { }
75 void __weak
arch_cpu_idle_exit(void) { }
76 void __weak
arch_cpu_idle_dead(void) { }
77 void __weak
arch_cpu_idle(void)
79 cpu_idle_force_poll
= 1;
84 * default_idle_call - Default CPU idle routine.
86 * To use when the cpuidle framework cannot be used.
88 void __cpuidle
default_idle_call(void)
90 if (current_clr_polling_and_test()) {
93 stop_critical_timings();
95 start_critical_timings();
99 static int call_cpuidle(struct cpuidle_driver
*drv
, struct cpuidle_device
*dev
,
103 * The idle task must be scheduled, it is pointless to go to idle, just
104 * update no idle residency and return.
106 if (current_clr_polling_and_test()) {
107 dev
->last_residency_ns
= 0;
113 * Enter the idle state previously returned by the governor decision.
114 * This function will block until an interrupt occurs and will take
115 * care of re-enabling the local interrupts
117 return cpuidle_enter(drv
, dev
, next_state
);
121 * cpuidle_idle_call - the main idle function
123 * NOTE: no locks or semaphores should be used here
125 * On archs that support TIF_POLLING_NRFLAG, is called with polling
126 * set, and it returns with polling set. If it ever stops polling, it
127 * must clear the polling bit.
129 static void cpuidle_idle_call(void)
131 struct cpuidle_device
*dev
= cpuidle_get_device();
132 struct cpuidle_driver
*drv
= cpuidle_get_cpu_driver(dev
);
133 int next_state
, entered_state
;
136 * Check if the idle task must be rescheduled. If it is the
137 * case, exit the function after re-enabling the local irq.
139 if (need_resched()) {
145 * The RCU framework needs to be told that we are entering an idle
146 * section, so no more rcu read side critical sections and one more
147 * step to the grace period
150 if (cpuidle_not_available(drv
, dev
)) {
151 tick_nohz_idle_stop_tick();
159 * Suspend-to-idle ("s2idle") is a system state in which all user space
160 * has been frozen, all I/O devices have been suspended and the only
161 * activity happens here and in interrupts (if any). In that case bypass
162 * the cpuidle governor and go stratight for the deepest idle state
163 * available. Possibly also suspend the local tick and the entire
164 * timekeeping to prevent timer interrupts from kicking us out of idle
165 * until a proper wakeup interrupt happens.
168 if (idle_should_enter_s2idle() || dev
->forced_idle_latency_limit_ns
) {
171 if (idle_should_enter_s2idle()) {
174 entered_state
= cpuidle_enter_s2idle(drv
, dev
);
175 if (entered_state
> 0) {
182 max_latency_ns
= U64_MAX
;
184 max_latency_ns
= dev
->forced_idle_latency_limit_ns
;
187 tick_nohz_idle_stop_tick();
190 next_state
= cpuidle_find_deepest_state(drv
, dev
, max_latency_ns
);
191 call_cpuidle(drv
, dev
, next_state
);
193 bool stop_tick
= true;
196 * Ask the cpuidle framework to choose a convenient idle state.
198 next_state
= cpuidle_select(drv
, dev
, &stop_tick
);
200 if (stop_tick
|| tick_nohz_tick_stopped())
201 tick_nohz_idle_stop_tick();
203 tick_nohz_idle_retain_tick();
207 entered_state
= call_cpuidle(drv
, dev
, next_state
);
209 * Give the governor an opportunity to reflect on the outcome
211 cpuidle_reflect(dev
, entered_state
);
215 __current_set_polling();
218 * It is up to the idle functions to reenable local interrupts
220 if (WARN_ON_ONCE(irqs_disabled()))
227 * Generic idle loop implementation
229 * Called with polling cleared.
231 static void do_idle(void)
233 int cpu
= smp_processor_id();
235 * If the arch has a polling bit, we maintain an invariant:
237 * Our polling bit is clear if we're not scheduled (i.e. if rq->curr !=
238 * rq->idle). This means that, if rq->idle has the polling bit set,
239 * then setting need_resched is guaranteed to cause the CPU to
243 __current_set_polling();
244 tick_nohz_idle_enter();
246 while (!need_resched()) {
251 if (cpu_is_offline(cpu
)) {
252 tick_nohz_idle_stop_tick();
253 cpuhp_report_idle_dead();
254 arch_cpu_idle_dead();
257 arch_cpu_idle_enter();
260 * In poll mode we reenable interrupts and spin. Also if we
261 * detected in the wakeup from idle path that the tick
262 * broadcast device expired for us, we don't want to go deep
263 * idle as we know that the IPI is going to arrive right away.
265 if (cpu_idle_force_poll
|| tick_check_broadcast_expired()) {
266 tick_nohz_idle_restart_tick();
271 arch_cpu_idle_exit();
275 * Since we fell out of the loop above, we know TIF_NEED_RESCHED must
276 * be set, propagate it into PREEMPT_NEED_RESCHED.
278 * This is required because for polling idle loops we will not have had
279 * an IPI to fold the state for us.
281 preempt_set_need_resched();
282 tick_nohz_idle_exit();
283 __current_clr_polling();
286 * We promise to call sched_ttwu_pending() and reschedule if
287 * need_resched() is set while polling is set. That means that clearing
288 * polling needs to be visible before doing these things.
290 smp_mb__after_atomic();
292 sched_ttwu_pending();
295 if (unlikely(klp_patch_pending(current
)))
296 klp_update_patch_state(current
);
299 bool cpu_in_idle(unsigned long pc
)
301 return pc
>= (unsigned long)__cpuidle_text_start
&&
302 pc
< (unsigned long)__cpuidle_text_end
;
306 struct hrtimer timer
;
310 static enum hrtimer_restart
idle_inject_timer_fn(struct hrtimer
*timer
)
312 struct idle_timer
*it
= container_of(timer
, struct idle_timer
, timer
);
314 WRITE_ONCE(it
->done
, 1);
315 set_tsk_need_resched(current
);
317 return HRTIMER_NORESTART
;
320 void play_idle_precise(u64 duration_ns
, u64 latency_ns
)
322 struct idle_timer it
;
325 * Only FIFO tasks can disable the tick since they don't need the forced
328 WARN_ON_ONCE(current
->policy
!= SCHED_FIFO
);
329 WARN_ON_ONCE(current
->nr_cpus_allowed
!= 1);
330 WARN_ON_ONCE(!(current
->flags
& PF_KTHREAD
));
331 WARN_ON_ONCE(!(current
->flags
& PF_NO_SETAFFINITY
));
332 WARN_ON_ONCE(!duration_ns
);
336 current
->flags
|= PF_IDLE
;
337 cpuidle_use_deepest_state(latency_ns
);
340 hrtimer_init_on_stack(&it
.timer
, CLOCK_MONOTONIC
, HRTIMER_MODE_REL
);
341 it
.timer
.function
= idle_inject_timer_fn
;
342 hrtimer_start(&it
.timer
, ns_to_ktime(duration_ns
),
343 HRTIMER_MODE_REL_PINNED
);
345 while (!READ_ONCE(it
.done
))
348 cpuidle_use_deepest_state(0);
349 current
->flags
&= ~PF_IDLE
;
351 preempt_fold_need_resched();
354 EXPORT_SYMBOL_GPL(play_idle_precise
);
356 void cpu_startup_entry(enum cpuhp_state state
)
358 arch_cpu_idle_prepare();
359 cpuhp_online_idle(state
);
365 * idle-task scheduling class.
370 select_task_rq_idle(struct task_struct
*p
, int cpu
, int sd_flag
, int flags
)
372 return task_cpu(p
); /* IDLE tasks as never migrated */
376 balance_idle(struct rq
*rq
, struct task_struct
*prev
, struct rq_flags
*rf
)
378 return WARN_ON_ONCE(1);
383 * Idle tasks are unconditionally rescheduled:
385 static void check_preempt_curr_idle(struct rq
*rq
, struct task_struct
*p
, int flags
)
390 static void put_prev_task_idle(struct rq
*rq
, struct task_struct
*prev
)
394 static void set_next_task_idle(struct rq
*rq
, struct task_struct
*next
, bool first
)
396 update_idle_core(rq
);
397 schedstat_inc(rq
->sched_goidle
);
400 struct task_struct
*pick_next_task_idle(struct rq
*rq
)
402 struct task_struct
*next
= rq
->idle
;
404 set_next_task_idle(rq
, next
, true);
410 * It is not legal to sleep in the idle task - print a warning
411 * message if some code attempts to do it:
414 dequeue_task_idle(struct rq
*rq
, struct task_struct
*p
, int flags
)
416 raw_spin_unlock_irq(&rq
->lock
);
417 printk(KERN_ERR
"bad: scheduling from the idle thread!\n");
419 raw_spin_lock_irq(&rq
->lock
);
423 * scheduler tick hitting a task of our scheduling class.
425 * NOTE: This function can be called remotely by the tick offload that
426 * goes along full dynticks. Therefore no local assumption can be made
427 * and everything must be accessed through the @rq and @curr passed in
430 static void task_tick_idle(struct rq
*rq
, struct task_struct
*curr
, int queued
)
434 static void switched_to_idle(struct rq
*rq
, struct task_struct
*p
)
440 prio_changed_idle(struct rq
*rq
, struct task_struct
*p
, int oldprio
)
445 static unsigned int get_rr_interval_idle(struct rq
*rq
, struct task_struct
*task
)
450 static void update_curr_idle(struct rq
*rq
)
455 * Simple, special scheduling class for the per-CPU idle tasks:
457 const struct sched_class idle_sched_class
= {
459 /* no enqueue/yield_task for idle tasks */
461 /* dequeue is not valid, we print a debug message there: */
462 .dequeue_task
= dequeue_task_idle
,
464 .check_preempt_curr
= check_preempt_curr_idle
,
466 .pick_next_task
= pick_next_task_idle
,
467 .put_prev_task
= put_prev_task_idle
,
468 .set_next_task
= set_next_task_idle
,
471 .balance
= balance_idle
,
472 .select_task_rq
= select_task_rq_idle
,
473 .set_cpus_allowed
= set_cpus_allowed_common
,
476 .task_tick
= task_tick_idle
,
478 .get_rr_interval
= get_rr_interval_idle
,
480 .prio_changed
= prio_changed_idle
,
481 .switched_to
= switched_to_idle
,
482 .update_curr
= update_curr_idle
,