2 * Generic entry point for the idle threads
4 #include <linux/sched.h>
6 #include <linux/cpuidle.h>
7 #include <linux/tick.h>
9 #include <linux/stackprotector.h>
10 #include <linux/suspend.h>
14 #include <trace/events/power.h>
19 * sched_idle_set_state - Record idle state for the current CPU.
20 * @idle_state: State to record.
22 void sched_idle_set_state(struct cpuidle_state *idle_state, int index)
24 idle_set_state(this_rq(), idle_state);
25 idle_set_state_idx(this_rq(), index);
28 static int __read_mostly cpu_idle_force_poll;
30 void cpu_idle_poll_ctrl(bool enable)
33 cpu_idle_force_poll++;
35 cpu_idle_force_poll--;
36 WARN_ON_ONCE(cpu_idle_force_poll < 0);
40 #ifdef CONFIG_GENERIC_IDLE_POLL_SETUP
41 static int __init cpu_idle_poll_setup(char *__unused)
43 cpu_idle_force_poll = 1;
46 __setup("nohlt", cpu_idle_poll_setup);
48 static int __init cpu_idle_nopoll_setup(char *__unused)
50 cpu_idle_force_poll = 0;
53 __setup("hlt", cpu_idle_nopoll_setup);
56 static inline int cpu_idle_poll(void)
59 trace_cpu_idle_rcuidle(0, smp_processor_id());
61 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());
71 /* Weak implementations for optional arch specific functions */
72 void __weak arch_cpu_idle_prepare(void) { }
73 void __weak arch_cpu_idle_enter(void) { }
74 void __weak arch_cpu_idle_exit(void) { }
75 void __weak arch_cpu_idle_dead(void) { }
76 void __weak arch_cpu_idle(void)
78 cpu_idle_force_poll = 1;
83 * default_idle_call - Default CPU idle routine.
85 * To use when the cpuidle framework cannot be used.
87 void default_idle_call(void)
89 if (current_clr_polling_and_test()) {
92 stop_critical_timings();
94 start_critical_timings();
98 static int call_cpuidle(struct cpuidle_driver *drv, struct cpuidle_device *dev,
101 /* Fall back to the default arch idle method on errors. */
102 if (next_state < 0) {
108 * The idle task must be scheduled, it is pointless to go to idle, just
109 * update no idle residency and return.
111 if (current_clr_polling_and_test()) {
112 dev->last_residency = 0;
118 * Enter the idle state previously returned by the governor decision.
119 * This function will block until an interrupt occurs and will take
120 * care of re-enabling the local interrupts
122 return cpuidle_enter(drv, dev, next_state);
126 * cpuidle_idle_call - the main idle function
128 * NOTE: no locks or semaphores should be used here
130 * On archs that support TIF_POLLING_NRFLAG, is called with polling
131 * set, and it returns with polling set. If it ever stops polling, it
132 * must clear the polling bit.
134 static void cpuidle_idle_call(void)
136 struct cpuidle_device *dev = __this_cpu_read(cpuidle_devices);
137 struct cpuidle_driver *drv = cpuidle_get_cpu_driver(dev);
138 int next_state, entered_state;
141 * Check if the idle task must be rescheduled. If it is the
142 * case, exit the function after re-enabling the local irq.
144 if (need_resched()) {
150 * Tell the RCU framework we are entering an idle section,
151 * so no more rcu read side critical sections and one more
152 * step to the grace period
156 if (cpuidle_not_available(drv, dev)) {
162 * Suspend-to-idle ("freeze") is a system state in which all user space
163 * has been frozen, all I/O devices have been suspended and the only
164 * activity happens here and in iterrupts (if any). In that case bypass
165 * the cpuidle governor and go stratight for the deepest idle state
166 * available. Possibly also suspend the local tick and the entire
167 * timekeeping to prevent timer interrupts from kicking us out of idle
168 * until a proper wakeup interrupt happens.
170 if (idle_should_freeze()) {
171 entered_state = cpuidle_enter_freeze(drv, dev);
172 if (entered_state >= 0) {
177 next_state = cpuidle_find_deepest_state(drv, dev);
178 call_cpuidle(drv, dev, next_state);
181 * Ask the cpuidle framework to choose a convenient idle state.
183 next_state = cpuidle_select(drv, dev);
184 entered_state = call_cpuidle(drv, dev, next_state);
186 * Give the governor an opportunity to reflect on the outcome
188 cpuidle_reflect(dev, entered_state);
192 __current_set_polling();
195 * It is up to the idle functions to reenable local interrupts
197 if (WARN_ON_ONCE(irqs_disabled()))
203 DEFINE_PER_CPU(bool, cpu_dead_idle);
206 * Generic idle loop implementation
208 * Called with polling cleared.
210 static void cpu_idle_loop(void)
214 * If the arch has a polling bit, we maintain an invariant:
216 * Our polling bit is clear if we're not scheduled (i.e. if
217 * rq->curr != rq->idle). This means that, if rq->idle has
218 * the polling bit set, then setting need_resched is
219 * guaranteed to cause the cpu to reschedule.
222 __current_set_polling();
224 tick_nohz_idle_enter();
226 while (!need_resched()) {
230 if (cpu_is_offline(smp_processor_id())) {
231 rcu_cpu_notify(NULL, CPU_DYING_IDLE,
232 (void *)(long)smp_processor_id());
233 smp_mb(); /* all activity before dead. */
234 this_cpu_write(cpu_dead_idle, true);
235 arch_cpu_idle_dead();
239 arch_cpu_idle_enter();
242 * In poll mode we reenable interrupts and spin.
244 * Also if we detected in the wakeup from idle
245 * path that the tick broadcast device expired
246 * for us, we don't want to go deep idle as we
247 * know that the IPI is going to arrive right
250 if (cpu_idle_force_poll || tick_check_broadcast_expired())
255 arch_cpu_idle_exit();
259 * Since we fell out of the loop above, we know
260 * TIF_NEED_RESCHED must be set, propagate it into
261 * PREEMPT_NEED_RESCHED.
263 * This is required because for polling idle loops we will
264 * not have had an IPI to fold the state for us.
266 preempt_set_need_resched();
267 tick_nohz_idle_exit();
268 __current_clr_polling();
271 * We promise to call sched_ttwu_pending and reschedule
272 * if need_resched is set while polling is set. That
273 * means that clearing polling needs to be visible
274 * before doing these things.
276 smp_mb__after_atomic();
278 sched_ttwu_pending();
279 schedule_preempt_disabled();
283 void cpu_startup_entry(enum cpuhp_state state)
286 * This #ifdef needs to die, but it's too late in the cycle to
287 * make this generic (arm and sh have never invoked the canary
288 * init for the non boot cpus!). Will be fixed in 3.11
292 * If we're the non-boot CPU, nothing set the stack canary up
293 * for us. The boot CPU already has it initialized but no harm
294 * in doing it again. This is a good place for updating it, as
295 * we wont ever return from this function (so the invalid
296 * canaries already on the stack wont ever trigger).
298 boot_init_stack_canary();
300 arch_cpu_idle_prepare();