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
10 #include <linux/suspend.h>
19  * sched_idle_set_state - Record idle state for the current CPU.
20  * @idle_state: State to record.
34 		cpu_idle_force_poll--;  in cpu_idle_poll_ctrl()
109  * default_idle_call - Default CPU idle routine.
111  * To use when the cpuidle framework cannot be used.
137 		return -EBUSY;  in call_cpuidle_s2idle()
146 	 * The idle task must be scheduled, it is pointless to go to idle, just  in call_cpuidle()
147 	 * update no idle residency and return.  in call_cpuidle()
150 		dev->last_residency_ns = 0;  in call_cpuidle()
152 		return -EBUSY;  in call_cpuidle()
156 	 * Enter the idle state previously returned by the governor decision.  in call_cpuidle()
158 	 * care of re-enabling the local interrupts  in call_cpuidle()
164  * cpuidle_idle_call - the main idle function
179 	 * Check if the idle task must be rescheduled. If it is the  in cpuidle_idle_call()
180 	 * case, exit the function after re-enabling the local IRQ.  in cpuidle_idle_call()
195 	 * Suspend-to-idle ("s2idle") is a system state in which all user space  in cpuidle_idle_call()
198 	 * the cpuidle governor and go straight for the deepest idle state  in cpuidle_idle_call()
199 	 * available.  Possibly also suspend the local tick and the entire  in cpuidle_idle_call()
200 	 * timekeeping to prevent timer interrupts from kicking us out of idle  in cpuidle_idle_call()
204 	if (idle_should_enter_s2idle() || dev->forced_idle_latency_limit_ns) {  in cpuidle_idle_call()
215 			max_latency_ns = dev->forced_idle_latency_limit_ns;  in cpuidle_idle_call()
226 		 * Ask the cpuidle framework to choose a convenient idle state.  in cpuidle_idle_call()
237 		 * Give the governor an opportunity to reflect on the outcome  in cpuidle_idle_call()
246 	 * It is up to the idle functions to re-enable local interrupts  in cpuidle_idle_call()
253  * Generic idle loop implementation
262 	 * Check if we need to update blocked load  in do_idle()
269 	 * Our polling bit is clear if we're not scheduled (i.e. if rq->curr !=  in do_idle()
270 	 * rq->idle). This means that, if rq->idle has the polling bit set,  in do_idle()
271 	 * then setting need_resched is guaranteed to cause the CPU to  in do_idle()
281 		 * Interrupts shouldn't be re-enabled from that point on until  in do_idle()
287 		 * Several cases to consider:  in do_idle()
289 		 * - SLEEP-UNTIL-PENDING-INTERRUPT based instructions such as  in do_idle()
293 		 * - sti;mwait() couple is fine because the interrupts are  in do_idle()
294 		 *   re-enabled only upon the execution of mwait, leaving no gap  in do_idle()
295 		 *   in-between.  in do_idle()
297 		 * - ROLLBACK based idle handlers with the sleeping instruction  in do_idle()
299 		 *   when the interrupt detects it has interrupted an idle handler,  in do_idle()
300 		 *   it rolls back to its beginning which performs the  in do_idle()
301 		 *   need_resched() check before re-executing the sleeping  in do_idle()
303 		 *   If such a scheme is really mandatory due to the lack of an  in do_idle()
304 		 *   appropriate CPU sleeping instruction, then a FAST-FORWARD  in do_idle()
306 		 *   interrupted an idle handler, it must resume to the end of  in do_idle()
307 		 *   this idle handler so that the generic idle loop is iterated  in do_idle()
308 		 *   again to reprogram the tick.  in do_idle()
321 		 * In poll mode we re-enable interrupts and spin. Also if we  in do_idle()
322 		 * detected in the wakeup from idle path that the tick  in do_idle()
323 		 * broadcast device expired for us, we don't want to go deep  in do_idle()
324 		 * idle as we know that the IPI is going to arrive right away.  in do_idle()
339 	 * This is required because for polling idle loops we will not have had  in do_idle()
340 	 * an IPI to fold the state for us.  in do_idle()
347 	 * We promise to call sched_ttwu_pending() and reschedule if  in do_idle()
349 	 * polling needs to be visible before doing these things.  in do_idle()
354 	 * RCU relies on this call to be done outside of an RCU read-side  in do_idle()
379 	WRITE_ONCE(it->done, 1);  in idle_inject_timer_fn()
393 	WARN_ON_ONCE(current->policy != SCHED_FIFO);  in play_idle_precise()
394 	WARN_ON_ONCE(current->nr_cpus_allowed != 1);  in play_idle_precise()
395 	WARN_ON_ONCE(!(current->flags & PF_KTHREAD));  in play_idle_precise()
396 	WARN_ON_ONCE(!(current->flags & PF_NO_SETAFFINITY));  in play_idle_precise()
398 	WARN_ON_ONCE(current->mm);  in play_idle_precise()
402 	current->flags |= PF_IDLE;  in play_idle_precise()
415 	current->flags &= ~PF_IDLE;  in play_idle_precise()
424 	current->flags |= PF_IDLE;  in cpu_startup_entry()
432  * idle-task scheduling class.
438 	return task_cpu(p); /* IDLE tasks as never migrated */  in select_task_rq_idle()
448  * Idle tasks are unconditionally rescheduled:
465 	schedstat_inc(rq->sched_goidle);  in set_next_task_idle()
466 	next->se.exec_start = rq_clock_task(rq);  in set_next_task_idle()
472 	return rq->idle;  in pick_task_idle()
476  * It is not legal to sleep in the idle task - print a warning
477  * message if some code attempts to do it:
483 	printk(KERN_ERR "bad: scheduling from the idle thread!\n");  in dequeue_task_idle()
517  * Simple, special scheduling class for the per-CPU idle tasks:
519 DEFINE_SCHED_CLASS(idle) = {
521 	/* no enqueue/yield_task for idle tasks */