diff options
Diffstat (limited to 'kernel/sched')
| -rw-r--r-- | kernel/sched/clock.c | 48 | ||||
| -rw-r--r-- | kernel/sched/core.c | 352 | ||||
| -rw-r--r-- | kernel/sched/cpuacct.c | 147 | ||||
| -rw-r--r-- | kernel/sched/deadline.c | 26 | ||||
| -rw-r--r-- | kernel/sched/debug.c | 10 | ||||
| -rw-r--r-- | kernel/sched/fair.c | 358 | ||||
| -rw-r--r-- | kernel/sched/idle_task.c | 2 | ||||
| -rw-r--r-- | kernel/sched/rt.c | 15 | ||||
| -rw-r--r-- | kernel/sched/sched.h | 129 | ||||
| -rw-r--r-- | kernel/sched/stop_task.c | 2 |
10 files changed, 660 insertions, 429 deletions
diff --git a/kernel/sched/clock.c b/kernel/sched/clock.c index fedb967a9841..e85a725e5c34 100644 --- a/kernel/sched/clock.c +++ b/kernel/sched/clock.c @@ -318,6 +318,7 @@ u64 sched_clock_cpu(int cpu) return clock; } +EXPORT_SYMBOL_GPL(sched_clock_cpu); void sched_clock_tick(void) { @@ -363,39 +364,6 @@ void sched_clock_idle_wakeup_event(u64 delta_ns) } EXPORT_SYMBOL_GPL(sched_clock_idle_wakeup_event); -/* - * As outlined at the top, provides a fast, high resolution, nanosecond - * time source that is monotonic per cpu argument and has bounded drift - * between cpus. - * - * ######################### BIG FAT WARNING ########################## - * # when comparing cpu_clock(i) to cpu_clock(j) for i != j, time can # - * # go backwards !! # - * #################################################################### - */ -u64 cpu_clock(int cpu) -{ - if (!sched_clock_stable()) - return sched_clock_cpu(cpu); - - return sched_clock(); -} - -/* - * Similar to cpu_clock() for the current cpu. Time will only be observed - * to be monotonic if care is taken to only compare timestampt taken on the - * same CPU. - * - * See cpu_clock(). - */ -u64 local_clock(void) -{ - if (!sched_clock_stable()) - return sched_clock_cpu(raw_smp_processor_id()); - - return sched_clock(); -} - #else /* CONFIG_HAVE_UNSTABLE_SCHED_CLOCK */ void sched_clock_init(void) @@ -410,22 +378,8 @@ u64 sched_clock_cpu(int cpu) return sched_clock(); } - -u64 cpu_clock(int cpu) -{ - return sched_clock(); -} - -u64 local_clock(void) -{ - return sched_clock(); -} - #endif /* CONFIG_HAVE_UNSTABLE_SCHED_CLOCK */ -EXPORT_SYMBOL_GPL(cpu_clock); -EXPORT_SYMBOL_GPL(local_clock); - /* * Running clock - returns the time that has elapsed while a guest has been * running. diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 9c710ad0ac22..1e622f254df4 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -33,7 +33,7 @@ #include <linux/init.h> #include <linux/uaccess.h> #include <linux/highmem.h> -#include <asm/mmu_context.h> +#include <linux/mmu_context.h> #include <linux/interrupt.h> #include <linux/capability.h> #include <linux/completion.h> @@ -170,6 +170,71 @@ static struct rq *this_rq_lock(void) return rq; } +/* + * __task_rq_lock - lock the rq @p resides on. + */ +struct rq *__task_rq_lock(struct task_struct *p, struct rq_flags *rf) + __acquires(rq->lock) +{ + struct rq *rq; + + lockdep_assert_held(&p->pi_lock); + + for (;;) { + rq = task_rq(p); + raw_spin_lock(&rq->lock); + if (likely(rq == task_rq(p) && !task_on_rq_migrating(p))) { + rf->cookie = lockdep_pin_lock(&rq->lock); + return rq; + } + raw_spin_unlock(&rq->lock); + + while (unlikely(task_on_rq_migrating(p))) + cpu_relax(); + } +} + +/* + * task_rq_lock - lock p->pi_lock and lock the rq @p resides on. + */ +struct rq *task_rq_lock(struct task_struct *p, struct rq_flags *rf) + __acquires(p->pi_lock) + __acquires(rq->lock) +{ + struct rq *rq; + + for (;;) { + raw_spin_lock_irqsave(&p->pi_lock, rf->flags); + rq = task_rq(p); + raw_spin_lock(&rq->lock); + /* + * move_queued_task() task_rq_lock() + * + * ACQUIRE (rq->lock) + * [S] ->on_rq = MIGRATING [L] rq = task_rq() + * WMB (__set_task_cpu()) ACQUIRE (rq->lock); + * [S] ->cpu = new_cpu [L] task_rq() + * [L] ->on_rq + * RELEASE (rq->lock) + * + * If we observe the old cpu in task_rq_lock, the acquire of + * the old rq->lock will fully serialize against the stores. + * + * If we observe the new cpu in task_rq_lock, the acquire will + * pair with the WMB to ensure we must then also see migrating. + */ + if (likely(rq == task_rq(p) && !task_on_rq_migrating(p))) { + rf->cookie = lockdep_pin_lock(&rq->lock); + return rq; + } + raw_spin_unlock(&rq->lock); + raw_spin_unlock_irqrestore(&p->pi_lock, rf->flags); + + while (unlikely(task_on_rq_migrating(p))) + cpu_relax(); + } +} + #ifdef CONFIG_SCHED_HRTICK /* * Use HR-timers to deliver accurate preemption points. @@ -369,7 +434,7 @@ void wake_q_add(struct wake_q_head *head, struct task_struct *task) * wakeup due to that. * * This cmpxchg() implies a full barrier, which pairs with the write - * barrier implied by the wakeup in wake_up_list(). + * barrier implied by the wakeup in wake_up_q(). */ if (cmpxchg(&node->next, NULL, WAKE_Q_TAIL)) return; @@ -565,17 +630,8 @@ bool sched_can_stop_tick(struct rq *rq) return false; /* - * FIFO realtime policy runs the highest priority task (after DEADLINE). - * Other runnable tasks are of a lower priority. The scheduler tick - * isn't needed. - */ - fifo_nr_running = rq->rt.rt_nr_running - rq->rt.rr_nr_running; - if (fifo_nr_running) - return true; - - /* - * Round-robin realtime tasks time slice with other tasks at the same - * realtime priority. + * If there are more than one RR tasks, we need the tick to effect the + * actual RR behaviour. */ if (rq->rt.rr_nr_running) { if (rq->rt.rr_nr_running == 1) @@ -584,8 +640,20 @@ bool sched_can_stop_tick(struct rq *rq) return false; } - /* Normal multitasking need periodic preemption checks */ - if (rq->cfs.nr_running > 1) + /* + * If there's no RR tasks, but FIFO tasks, we can skip the tick, no + * forced preemption between FIFO tasks. + */ + fifo_nr_running = rq->rt.rt_nr_running - rq->rt.rr_nr_running; + if (fifo_nr_running) + return true; + + /* + * If there are no DL,RR/FIFO tasks, there must only be CFS tasks left; + * if there's more than one we need the tick for involuntary + * preemption. + */ + if (rq->nr_running > 1) return false; return true; @@ -1053,11 +1121,11 @@ static int __set_cpus_allowed_ptr(struct task_struct *p, { const struct cpumask *cpu_valid_mask = cpu_active_mask; unsigned int dest_cpu; - unsigned long flags; + struct rq_flags rf; struct rq *rq; int ret = 0; - rq = task_rq_lock(p, &flags); + rq = task_rq_lock(p, &rf); if (p->flags & PF_KTHREAD) { /* @@ -1103,7 +1171,7 @@ static int __set_cpus_allowed_ptr(struct task_struct *p, if (task_running(rq, p) || p->state == TASK_WAKING) { struct migration_arg arg = { p, dest_cpu }; /* Need help from migration thread: drop lock and wait. */ - task_rq_unlock(rq, p, &flags); + task_rq_unlock(rq, p, &rf); stop_one_cpu(cpu_of(rq), migration_cpu_stop, &arg); tlb_migrate_finish(p->mm); return 0; @@ -1112,12 +1180,12 @@ static int __set_cpus_allowed_ptr(struct task_struct *p, * OK, since we're going to drop the lock immediately * afterwards anyway. */ - lockdep_unpin_lock(&rq->lock); + lockdep_unpin_lock(&rq->lock, rf.cookie); rq = move_queued_task(rq, p, dest_cpu); - lockdep_pin_lock(&rq->lock); + lockdep_repin_lock(&rq->lock, rf.cookie); } out: - task_rq_unlock(rq, p, &flags); + task_rq_unlock(rq, p, &rf); return ret; } @@ -1301,8 +1369,8 @@ out: */ unsigned long wait_task_inactive(struct task_struct *p, long match_state) { - unsigned long flags; int running, queued; + struct rq_flags rf; unsigned long ncsw; struct rq *rq; @@ -1337,14 +1405,14 @@ unsigned long wait_task_inactive(struct task_struct *p, long match_state) * lock now, to be *sure*. If we're wrong, we'll * just go back and repeat. */ - rq = task_rq_lock(p, &flags); + rq = task_rq_lock(p, &rf); trace_sched_wait_task(p); running = task_running(rq, p); queued = task_on_rq_queued(p); ncsw = 0; if (!match_state || p->state == match_state) ncsw = p->nvcsw | LONG_MIN; /* sets MSB */ - task_rq_unlock(rq, p, &flags); + task_rq_unlock(rq, p, &rf); /* * If it changed from the expected state, bail out now. @@ -1605,8 +1673,8 @@ static inline void ttwu_activate(struct rq *rq, struct task_struct *p, int en_fl /* * Mark the task runnable and perform wakeup-preemption. */ -static void -ttwu_do_wakeup(struct rq *rq, struct task_struct *p, int wake_flags) +static void ttwu_do_wakeup(struct rq *rq, struct task_struct *p, int wake_flags, + struct pin_cookie cookie) { check_preempt_curr(rq, p, wake_flags); p->state = TASK_RUNNING; @@ -1618,9 +1686,9 @@ ttwu_do_wakeup(struct rq *rq, struct task_struct *p, int wake_flags) * Our task @p is fully woken up and running; so its safe to * drop the rq->lock, hereafter rq is only used for statistics. */ - lockdep_unpin_lock(&rq->lock); + lockdep_unpin_lock(&rq->lock, cookie); p->sched_class->task_woken(rq, p); - lockdep_pin_lock(&rq->lock); + lockdep_repin_lock(&rq->lock, cookie); } if (rq->idle_stamp) { @@ -1638,7 +1706,8 @@ ttwu_do_wakeup(struct rq *rq, struct task_struct *p, int wake_flags) } static void -ttwu_do_activate(struct rq *rq, struct task_struct *p, int wake_flags) +ttwu_do_activate(struct rq *rq, struct task_struct *p, int wake_flags, + struct pin_cookie cookie) { lockdep_assert_held(&rq->lock); @@ -1648,7 +1717,7 @@ ttwu_do_activate(struct rq *rq, struct task_struct *p, int wake_flags) #endif ttwu_activate(rq, p, ENQUEUE_WAKEUP | ENQUEUE_WAKING); - ttwu_do_wakeup(rq, p, wake_flags); + ttwu_do_wakeup(rq, p, wake_flags, cookie); } /* @@ -1659,17 +1728,18 @@ ttwu_do_activate(struct rq *rq, struct task_struct *p, int wake_flags) */ static int ttwu_remote(struct task_struct *p, int wake_flags) { + struct rq_flags rf; struct rq *rq; int ret = 0; - rq = __task_rq_lock(p); + rq = __task_rq_lock(p, &rf); if (task_on_rq_queued(p)) { /* check_preempt_curr() may use rq clock */ update_rq_clock(rq); - ttwu_do_wakeup(rq, p, wake_flags); + ttwu_do_wakeup(rq, p, wake_flags, rf.cookie); ret = 1; } - __task_rq_unlock(rq); + __task_rq_unlock(rq, &rf); return ret; } @@ -1679,6 +1749,7 @@ void sched_ttwu_pending(void) { struct rq *rq = this_rq(); struct llist_node *llist = llist_del_all(&rq->wake_list); + struct pin_cookie cookie; struct task_struct *p; unsigned long flags; @@ -1686,15 +1757,15 @@ void sched_ttwu_pending(void) return; raw_spin_lock_irqsave(&rq->lock, flags); - lockdep_pin_lock(&rq->lock); + cookie = lockdep_pin_lock(&rq->lock); while (llist) { p = llist_entry(llist, struct task_struct, wake_entry); llist = llist_next(llist); - ttwu_do_activate(rq, p, 0); + ttwu_do_activate(rq, p, 0, cookie); } - lockdep_unpin_lock(&rq->lock); + lockdep_unpin_lock(&rq->lock, cookie); raw_spin_unlock_irqrestore(&rq->lock, flags); } @@ -1781,6 +1852,7 @@ bool cpus_share_cache(int this_cpu, int that_cpu) static void ttwu_queue(struct task_struct *p, int cpu) { struct rq *rq = cpu_rq(cpu); + struct pin_cookie cookie; #if defined(CONFIG_SMP) if (sched_feat(TTWU_QUEUE) && !cpus_share_cache(smp_processor_id(), cpu)) { @@ -1791,9 +1863,9 @@ static void ttwu_queue(struct task_struct *p, int cpu) #endif raw_spin_lock(&rq->lock); - lockdep_pin_lock(&rq->lock); - ttwu_do_activate(rq, p, 0); - lockdep_unpin_lock(&rq->lock); + cookie = lockdep_pin_lock(&rq->lock); + ttwu_do_activate(rq, p, 0, cookie); + lockdep_unpin_lock(&rq->lock, cookie); raw_spin_unlock(&rq->lock); } @@ -1990,7 +2062,7 @@ out: * ensure that this_rq() is locked, @p is bound to this_rq() and not * the current task. */ -static void try_to_wake_up_local(struct task_struct *p) +static void try_to_wake_up_local(struct task_struct *p, struct pin_cookie cookie) { struct rq *rq = task_rq(p); @@ -2007,11 +2079,11 @@ static void try_to_wake_up_local(struct task_struct *p) * disabled avoiding further scheduler activity on it and we've * not yet picked a replacement task. */ - lockdep_unpin_lock(&rq->lock); + lockdep_unpin_lock(&rq->lock, cookie); raw_spin_unlock(&rq->lock); raw_spin_lock(&p->pi_lock); raw_spin_lock(&rq->lock); - lockdep_pin_lock(&rq->lock); + lockdep_repin_lock(&rq->lock, cookie); } if (!(p->state & TASK_NORMAL)) @@ -2022,7 +2094,7 @@ static void try_to_wake_up_local(struct task_struct *p) if (!task_on_rq_queued(p)) ttwu_activate(rq, p, ENQUEUE_WAKEUP); - ttwu_do_wakeup(rq, p, 0); + ttwu_do_wakeup(rq, p, 0, cookie); if (schedstat_enabled()) ttwu_stat(p, smp_processor_id(), 0); out: @@ -2382,7 +2454,8 @@ static int dl_overflow(struct task_struct *p, int policy, u64 new_bw = dl_policy(policy) ? to_ratio(period, runtime) : 0; int cpus, err = -1; - if (new_bw == p->dl.dl_bw) + /* !deadline task may carry old deadline bandwidth */ + if (new_bw == p->dl.dl_bw && task_has_dl_policy(p)) return 0; /* @@ -2421,12 +2494,12 @@ extern void init_dl_bw(struct dl_bw *dl_b); */ void wake_up_new_task(struct task_struct *p) { - unsigned long flags; + struct rq_flags rf; struct rq *rq; - raw_spin_lock_irqsave(&p->pi_lock, flags); /* Initialize new task's runnable average */ init_entity_runnable_average(&p->se); + raw_spin_lock_irqsave(&p->pi_lock, rf.flags); #ifdef CONFIG_SMP /* * Fork balancing, do it here and not earlier because: @@ -2435,8 +2508,10 @@ void wake_up_new_task(struct task_struct *p) */ set_task_cpu(p, select_task_rq(p, task_cpu(p), SD_BALANCE_FORK, 0)); #endif + /* Post initialize new task's util average when its cfs_rq is set */ + post_init_entity_util_avg(&p->se); - rq = __task_rq_lock(p); + rq = __task_rq_lock(p, &rf); activate_task(rq, p, 0); p->on_rq = TASK_ON_RQ_QUEUED; trace_sched_wakeup_new(p); @@ -2447,12 +2522,12 @@ void wake_up_new_task(struct task_struct *p) * Nothing relies on rq->lock after this, so its fine to * drop it. */ - lockdep_unpin_lock(&rq->lock); + lockdep_unpin_lock(&rq->lock, rf.cookie); p->sched_class->task_woken(rq, p); - lockdep_pin_lock(&rq->lock); + lockdep_repin_lock(&rq->lock, rf.cookie); } #endif - task_rq_unlock(rq, p, &flags); + task_rq_unlock(rq, p, &rf); } #ifdef CONFIG_PREEMPT_NOTIFIERS @@ -2714,7 +2789,7 @@ asmlinkage __visible void schedule_tail(struct task_struct *prev) */ static __always_inline struct rq * context_switch(struct rq *rq, struct task_struct *prev, - struct task_struct *next) + struct task_struct *next, struct pin_cookie cookie) { struct mm_struct *mm, *oldmm; @@ -2734,7 +2809,7 @@ context_switch(struct rq *rq, struct task_struct *prev, atomic_inc(&oldmm->mm_count); enter_lazy_tlb(oldmm, next); } else - switch_mm(oldmm, mm, next); + switch_mm_irqs_off(oldmm, mm, next); if (!prev->mm) { prev->active_mm = NULL; @@ -2746,7 +2821,7 @@ context_switch(struct rq *rq, struct task_struct *prev, * of the scheduler it's an obvious special-case), so we * do an early lockdep release here: */ - lockdep_unpin_lock(&rq->lock); + lockdep_unpin_lock(&rq->lock, cookie); spin_release(&rq->lock.dep_map, 1, _THIS_IP_); /* Here we just switch the register state and the stack. */ @@ -2868,7 +2943,7 @@ EXPORT_PER_CPU_SYMBOL(kernel_cpustat); */ unsigned long long task_sched_runtime(struct task_struct *p) { - unsigned long flags; + struct rq_flags rf; struct rq *rq; u64 ns; @@ -2888,7 +2963,7 @@ unsigned long long task_sched_runtime(struct task_struct *p) return p->se.sum_exec_runtime; #endif - rq = task_rq_lock(p, &flags); + rq = task_rq_lock(p, &rf); /* * Must be ->curr _and_ ->on_rq. If dequeued, we would * project cycles that may never be accounted to this @@ -2899,7 +2974,7 @@ unsigned long long task_sched_runtime(struct task_struct *p) p->sched_class->update_curr(rq); } ns = p->se.sum_exec_runtime; - task_rq_unlock(rq, p, &flags); + task_rq_unlock(rq, p, &rf); return ns; } @@ -2919,7 +2994,7 @@ void scheduler_tick(void) raw_spin_lock(&rq->lock); update_rq_clock(rq); curr->sched_class->task_tick(rq, curr, 0); - update_cpu_load_active(rq); + cpu_load_update_active(rq); calc_global_load_tick(rq); raw_spin_unlock(&rq->lock); @@ -2962,6 +3037,20 @@ u64 scheduler_tick_max_deferment(void) #if defined(CONFIG_PREEMPT) && (defined(CONFIG_DEBUG_PREEMPT) || \ defined(CONFIG_PREEMPT_TRACER)) +/* + * If the value passed in is equal to the current preempt count + * then we just disabled preemption. Start timing the latency. + */ +static inline void preempt_latency_start(int val) +{ + if (preempt_count() == val) { + unsigned long ip = get_lock_parent_ip(); +#ifdef CONFIG_DEBUG_PREEMPT + current->preempt_disable_ip = ip; +#endif + trace_preempt_off(CALLER_ADDR0, ip); + } +} void preempt_count_add(int val) { @@ -2980,17 +3069,21 @@ void preempt_count_add(int val) DEBUG_LOCKS_WARN_ON((preempt_count() & PREEMPT_MASK) >= PREEMPT_MASK - 10); #endif - if (preempt_count() == val) { - unsigned long ip = get_lock_parent_ip(); -#ifdef CONFIG_DEBUG_PREEMPT - current->preempt_disable_ip = ip; -#endif - trace_preempt_off(CALLER_ADDR0, ip); - } + preempt_latency_start(val); } EXPORT_SYMBOL(preempt_count_add); NOKPROBE_SYMBOL(preempt_count_add); +/* + * If the value passed in equals to the current preempt count + * then we just enabled preemption. Stop timing the latency. + */ +static inline void preempt_latency_stop(int val) +{ + if (preempt_count() == val) + trace_preempt_on(CALLER_ADDR0, get_lock_parent_ip()); +} + void preempt_count_sub(int val) { #ifdef CONFIG_DEBUG_PREEMPT @@ -3007,13 +3100,15 @@ void preempt_count_sub(int val) return; #endif - if (preempt_count() == val) - trace_preempt_on(CALLER_ADDR0, get_lock_parent_ip()); + preempt_latency_stop(val); __preempt_count_sub(val); } EXPORT_SYMBOL(preempt_count_sub); NOKPROBE_SYMBOL(preempt_count_sub); +#else +static inline void preempt_latency_start(int val) { } +static inline void preempt_latency_stop(int val) { } #endif /* @@ -3066,7 +3161,7 @@ static inline void schedule_debug(struct task_struct *prev) * Pick up the highest-prio task: */ static inline struct task_struct * -pick_next_task(struct rq *rq, struct task_struct *prev) +pick_next_task(struct rq *rq, struct task_struct *prev, struct pin_cookie cookie) { const struct sched_class *class = &fair_sched_class; struct task_struct *p; @@ -3077,20 +3172,20 @@ pick_next_task(struct rq *rq, struct task_struct *prev) */ if (likely(prev->sched_class == class && rq->nr_running == rq->cfs.h_nr_running)) { - p = fair_sched_class.pick_next_task(rq, prev); + p = fair_sched_class.pick_next_task(rq, prev, cookie); if (unlikely(p == RETRY_TASK)) goto again; /* assumes fair_sched_class->next == idle_sched_class */ if (unlikely(!p)) - p = idle_sched_class.pick_next_task(rq, prev); + p = idle_sched_class.pick_next_task(rq, prev, cookie); return p; } again: for_each_class(class) { - p = class->pick_next_task(rq, prev); + p = class->pick_next_task(rq, prev, cookie); if (p) { if (unlikely(p == RETRY_TASK)) goto again; @@ -3144,6 +3239,7 @@ static void __sched notrace __schedule(bool preempt) { struct task_struct *prev, *next; unsigned long *switch_count; + struct pin_cookie cookie; struct rq *rq; int cpu; @@ -3177,7 +3273,7 @@ static void __sched notrace __schedule(bool preempt) */ smp_mb__before_spinlock(); raw_spin_lock(&rq->lock); - lockdep_pin_lock(&rq->lock); + cookie = lockdep_pin_lock(&rq->lock); rq->clock_skip_update <<= 1; /* promote REQ to ACT */ @@ -3199,7 +3295,7 @@ static void __sched notrace __schedule(bool preempt) to_wakeup = wq_worker_sleeping(prev); if (to_wakeup) - try_to_wake_up_local(to_wakeup); + try_to_wake_up_local(to_wakeup, cookie); } } switch_count = &prev->nvcsw; @@ -3208,7 +3304,7 @@ static void __sched notrace __schedule(bool preempt) if (task_on_rq_queued(prev)) update_rq_clock(rq); - next = pick_next_task(rq, prev); + next = pick_next_task(rq, prev, cookie); clear_tsk_need_resched(prev); clear_preempt_need_resched(); rq->clock_skip_update = 0; @@ -3219,9 +3315,9 @@ static void __sched notrace __schedule(bool preempt) ++*switch_count; trace_sched_switch(preempt, prev, next); - rq = context_switch(rq, prev, next); /* unlocks the rq */ + rq = context_switch(rq, prev, next, cookie); /* unlocks the rq */ } else { - lockdep_unpin_lock(&rq->lock); + lockdep_unpin_lock(&rq->lock, cookie); raw_spin_unlock_irq(&rq->lock); } @@ -3288,8 +3384,23 @@ void __sched schedule_preempt_disabled(void) static void __sched notrace preempt_schedule_common(void) { do { + /* + * Because the function tracer can trace preempt_count_sub() + * and it also uses preempt_enable/disable_notrace(), if + * NEED_RESCHED is set, the preempt_enable_notrace() called + * by the function tracer will call this function again and + * cause infinite recursion. + * + * Preemption must be disabled here before the function + * tracer can trace. Break up preempt_disable() into two + * calls. One to disable preemption without fear of being + * traced. The other to still record the preemption latency, + * which can also be traced by the function tracer. + */ preempt_disable_notrace(); + preempt_latency_start(1); __schedule(true); + preempt_latency_stop(1); preempt_enable_no_resched_notrace(); /* @@ -3341,7 +3452,21 @@ asmlinkage __visible void __sched notrace preempt_schedule_notrace(void) return; do { + /* + * Because the function tracer can trace preempt_count_sub() + * and it also uses preempt_enable/disable_notrace(), if + * NEED_RESCHED is set, the preempt_enable_notrace() called + * by the function tracer will call this function again and + * cause infinite recursion. + * + * Preemption must be disabled here before the function + * tracer can trace. Break up preempt_disable() into two + * calls. One to disable preemption without fear of being + * traced. The other to still record the preemption latency, + * which can also be traced by the function tracer. + */ preempt_disable_notrace(); + preempt_latency_start(1); /* * Needs preempt disabled in case user_exit() is traced * and the tracer calls preempt_enable_notrace() causing @@ -3351,6 +3476,7 @@ asmlinkage __visible void __sched notrace preempt_schedule_notrace(void) __schedule(true); exception_exit(prev_ctx); + preempt_latency_stop(1); preempt_enable_no_resched_notrace(); } while (need_resched()); } @@ -3407,12 +3533,13 @@ EXPORT_SYMBOL(default_wake_function); void rt_mutex_setprio(struct task_struct *p, int prio) { int oldprio, queued, running, queue_flag = DEQUEUE_SAVE | DEQUEUE_MOVE; - struct rq *rq; const struct sched_class *prev_class; + struct rq_flags rf; + struct rq *rq; BUG_ON(prio > MAX_PRIO); - rq = __task_rq_lock(p); + rq = __task_rq_lock(p, &rf); /* * Idle task boosting is a nono in general. There is one @@ -3488,7 +3615,7 @@ void rt_mutex_setprio(struct task_struct *p, int prio) check_class_changed(rq, p, prev_class, oldprio); out_unlock: preempt_disable(); /* avoid rq from going away on us */ - __task_rq_unlock(rq); + __task_rq_unlock(rq, &rf); balance_callback(rq); preempt_enable(); @@ -3498,7 +3625,7 @@ out_unlock: void set_user_nice(struct task_struct *p, long nice) { int old_prio, delta, queued; - unsigned long flags; + struct rq_flags rf; struct rq *rq; if (task_nice(p) == nice || nice < MIN_NICE || nice > MAX_NICE) @@ -3507,7 +3634,7 @@ void set_user_nice(struct task_struct *p, long nice) * We have to be careful, if called from sys_setpriority(), * the task might be in the middle of scheduling on another CPU. */ - rq = task_rq_lock(p, &flags); + rq = task_rq_lock(p, &rf); /* * The RT priorities are set via sched_setscheduler(), but we still * allow the 'normal' nice value to be set - but as expected @@ -3538,7 +3665,7 @@ void set_user_nice(struct task_struct *p, long nice) resched_curr(rq); } out_unlock: - task_rq_unlock(rq, p, &flags); + task_rq_unlock(rq, p, &rf); } EXPORT_SYMBOL(set_user_nice); @@ -3835,11 +3962,11 @@ static int __sched_setscheduler(struct task_struct *p, MAX_RT_PRIO - 1 - attr->sched_priority; int retval, oldprio, oldpolicy = -1, queued, running; int new_effective_prio, policy = attr->sched_policy; - unsigned long flags; const struct sched_class *prev_class; - struct rq *rq; + struct rq_flags rf; int reset_on_fork; int queue_flags = DEQUEUE_SAVE | DEQUEUE_MOVE; + struct rq *rq; /* may grab non-irq protected spin_locks */ BUG_ON(in_interrupt()); @@ -3934,13 +4061,13 @@ recheck: * To be able to change p->policy safely, the appropriate * runqueue lock must be held. */ - rq = task_rq_lock(p, &flags); + rq = task_rq_lock(p, &rf); /* * Changing the policy of the stop threads its a very bad idea */ if (p == rq->stop) { - task_rq_unlock(rq, p, &flags); + task_rq_unlock(rq, p, &rf); return -EINVAL; } @@ -3957,7 +4084,7 @@ recheck: goto change; p->sched_reset_on_fork = reset_on_fork; - task_rq_unlock(rq, p, &flags); + task_rq_unlock(rq, p, &rf); return 0; } change: @@ -3971,7 +4098,7 @@ change: if (rt_bandwidth_enabled() && rt_policy(policy) && task_group(p)->rt_bandwidth.rt_runtime == 0 && !task_group_is_autogroup(task_group(p))) { - task_rq_unlock(rq, p, &flags); + task_rq_unlock(rq, p, &rf); return -EPERM; } #endif @@ -3986,7 +4113,7 @@ change: */ if (!cpumask_subset(span, &p->cpus_allowed) || rq->rd->dl_bw.bw == 0) { - task_rq_unlock(rq, p, &flags); + task_rq_unlock(rq, p, &rf); return -EPERM; } } @@ -3996,7 +4123,7 @@ change: /* recheck policy now with rq lock held */ if (unlikely(oldpolicy != -1 && oldpolicy != p->policy)) { policy = oldpolicy = -1; - task_rq_unlock(rq, p, &flags); + task_rq_unlock(rq, p, &rf); goto recheck; } @@ -4006,7 +4133,7 @@ change: * is available. */ if ((dl_policy(policy) || dl_task(p)) && dl_overflow(p, policy, attr)) { - task_rq_unlock(rq, p, &flags); + task_rq_unlock(rq, p, &rf); return -EBUSY; } @@ -4051,7 +4178,7 @@ change: check_class_changed(rq, p, prev_class, oldprio); preempt_disable(); /* avoid rq from going away on us */ - task_rq_unlock(rq, p, &flags); + task_rq_unlock(rq, p, &rf); if (pi) rt_mutex_adjust_pi(p); @@ -4904,10 +5031,10 @@ SYSCALL_DEFINE2(sched_rr_get_interval, pid_t, pid, { struct task_struct *p; unsigned int time_slice; - unsigned long flags; + struct rq_flags rf; + struct timespec t; struct rq *rq; int retval; - struct timespec t; if (pid < 0) return -EINVAL; @@ -4922,11 +5049,11 @@ SYSCALL_DEFINE2(sched_rr_get_interval, pid_t, pid, if (retval) goto out_unlock; - rq = task_rq_lock(p, &flags); + rq = task_rq_lock(p, &rf); time_slice = 0; if (p->sched_class->get_rr_interval) time_slice = p->sched_class->get_rr_interval(rq, p); - task_rq_unlock(rq, p, &flags); + task_rq_unlock(rq, p, &rf); rcu_read_unlock(); jiffies_to_timespec(time_slice, &t); @@ -5002,7 +5129,8 @@ void show_state_filter(unsigned long state_filter) touch_all_softlockup_watchdogs(); #ifdef CONFIG_SCHED_DEBUG - sysrq_sched_debug_show(); + if (!state_filter) + sysrq_sched_debug_show(); #endif rcu_read_unlock(); /* @@ -5191,11 +5319,11 @@ int migrate_task_to(struct task_struct *p, int target_cpu) */ void sched_setnuma(struct task_struct *p, int nid) { - struct rq *rq; - unsigned long flags; bool queued, running; + struct rq_flags rf; + struct rq *rq; - rq = task_rq_lock(p, &flags); + rq = task_rq_lock(p, &rf); queued = task_on_rq_queued(p); running = task_current(rq, p); @@ -5210,7 +5338,7 @@ void sched_setnuma(struct task_struct *p, int nid) p->sched_class->set_curr_task(rq); if (queued) enqueue_task(rq, p, ENQUEUE_RESTORE); - task_rq_unlock(rq, p, &flags); + task_rq_unlock(rq, p, &rf); } #endif /* CONFIG_NUMA_BALANCING */ @@ -5226,7 +5354,7 @@ void idle_task_exit(void) BUG_ON(cpu_online(smp_processor_id())); if (mm != &init_mm) { - switch_mm(mm, &init_mm, current); + switch_mm_irqs_off(mm, &init_mm, current); finish_arch_post_lock_switch(); } mmdrop(mm); @@ -5274,6 +5402,7 @@ static void migrate_tasks(struct rq *dead_rq) { struct rq *rq = dead_rq; struct task_struct *next, *stop = rq->stop; + struct pin_cookie cookie; int dest_cpu; /* @@ -5305,8 +5434,8 @@ static void migrate_tasks(struct rq *dead_rq) /* * pick_next_task assumes pinned rq->lock. */ - lockdep_pin_lock(&rq->lock); - next = pick_next_task(rq, &fake_task); + cookie = lockdep_pin_lock(&rq->lock); + next = pick_next_task(rq, &fake_task, cookie); BUG_ON(!next); next->sched_class->put_prev_task(rq, next); @@ -5319,7 +5448,7 @@ static void migrate_tasks(struct rq *dead_rq) * because !cpu_active at this point, which means load-balance * will not interfere. Also, stop-machine. */ - lockdep_unpin_lock(&rq->lock); + lockdep_unpin_lock(&rq->lock, cookie); raw_spin_unlock(&rq->lock); raw_spin_lock(&next->pi_lock); raw_spin_lock(&rq->lock); @@ -7279,8 +7408,6 @@ void __init sched_init(void) for (j = 0; j < CPU_LOAD_IDX_MAX; j++) rq->cpu_load[j] = 0; - rq->last_load_update_tick = jiffies; - #ifdef CONFIG_SMP rq->sd = NULL; rq->rd = NULL; @@ -7299,12 +7426,13 @@ void __init sched_init(void) rq_attach_root(rq, &def_root_domain); #ifdef CONFIG_NO_HZ_COMMON + rq->last_load_update_tick = jiffies; rq->nohz_flags = 0; #endif #ifdef CONFIG_NO_HZ_FULL rq->last_sched_tick = 0; #endif -#endif +#endif /* CONFIG_SMP */ init_rq_hrtick(rq); atomic_set(&rq->nr_iowait, 0); } @@ -7587,10 +7715,10 @@ void sched_move_task(struct task_struct *tsk) { struct task_group *tg; int queued, running; - unsigned long flags; + struct rq_flags rf; struct rq *rq; - rq = task_rq_lock(tsk, &flags); + rq = task_rq_lock(tsk, &rf); running = task_current(rq, tsk); queued = task_on_rq_queued(tsk); @@ -7622,7 +7750,7 @@ void sched_move_task(struct task_struct *tsk) if (queued) enqueue_task(rq, tsk, ENQUEUE_RESTORE | ENQUEUE_MOVE); - task_rq_unlock(rq, tsk, &flags); + task_rq_unlock(rq, tsk, &rf); } #endif /* CONFIG_CGROUP_SCHED */ @@ -7842,7 +7970,7 @@ static int sched_rt_can_attach(struct task_group *tg, struct task_struct *tsk) static int sched_rt_global_constraints(void) { unsigned long flags; - int i, ret = 0; + int i; raw_spin_lock_irqsave(&def_rt_bandwidth.rt_runtime_lock, flags); for_each_possible_cpu(i) { @@ -7854,7 +7982,7 @@ static int sched_rt_global_constraints(void) } raw_spin_unlock_irqrestore(&def_rt_bandwidth.rt_runtime_lock, flags); - return ret; + return 0; } #endif /* CONFIG_RT_GROUP_SCHED */ diff --git a/kernel/sched/cpuacct.c b/kernel/sched/cpuacct.c index 4a811203c04a..41f85c4d0938 100644 --- a/kernel/sched/cpuacct.c +++ b/kernel/sched/cpuacct.c @@ -25,11 +25,22 @@ enum cpuacct_stat_index { CPUACCT_STAT_NSTATS, }; +enum cpuacct_usage_index { + CPUACCT_USAGE_USER, /* ... user mode */ + CPUACCT_USAGE_SYSTEM, /* ... kernel mode */ + + CPUACCT_USAGE_NRUSAGE, +}; + +struct cpuacct_usage { + u64 usages[CPUACCT_USAGE_NRUSAGE]; +}; + /* track cpu usage of a group of tasks and its child groups */ struct cpuacct { struct cgroup_subsys_state css; /* cpuusage holds pointer to a u64-type object on every cpu */ - u64 __percpu *cpuusage; + struct cpuacct_usage __percpu *cpuusage; struct kernel_cpustat __percpu *cpustat; }; @@ -49,7 +60,7 @@ static inline struct cpuacct *parent_ca(struct cpuacct *ca) return css_ca(ca->css.parent); } -static DEFINE_PER_CPU(u64, root_cpuacct_cpuusage); +static DEFINE_PER_CPU(struct cpuacct_usage, root_cpuacct_cpuusage); static struct cpuacct root_cpuacct = { .cpustat = &kernel_cpustat, .cpuusage = &root_cpuacct_cpuusage, @@ -68,7 +79,7 @@ cpuacct_css_alloc(struct cgroup_subsys_state *parent_css) if (!ca) goto out; - ca->cpuusage = alloc_percpu(u64); + ca->cpuusage = alloc_percpu(struct cpuacct_usage); if (!ca->cpuusage) goto out_free_ca; @@ -96,20 +107,37 @@ static void cpuacct_css_free(struct cgroup_subsys_state *css) kfree(ca); } -static u64 cpuacct_cpuusage_read(struct cpuacct *ca, int cpu) +static u64 cpuacct_cpuusage_read(struct cpuacct *ca, int cpu, + enum cpuacct_usage_index index) { - u64 *cpuusage = per_cpu_ptr(ca->cpuusage, cpu); + struct cpuacct_usage *cpuusage = per_cpu_ptr(ca->cpuusage, cpu); u64 data; + /* + * We allow index == CPUACCT_USAGE_NRUSAGE here to read + * the sum of suages. + */ + BUG_ON(index > CPUACCT_USAGE_NRUSAGE); + #ifndef CONFIG_64BIT /* * Take rq->lock to make 64-bit read safe on 32-bit platforms. */ raw_spin_lock_irq(&cpu_rq(cpu)->lock); - data = *cpuusage; +#endif + + if (index == CPUACCT_USAGE_NRUSAGE) { + int i = 0; + + data = 0; + for (i = 0; i < CPUACCT_USAGE_NRUSAGE; i++) + data += cpuusage->usages[i]; + } else { + data = cpuusage->usages[index]; + } + +#ifndef CONFIG_64BIT raw_spin_unlock_irq(&cpu_rq(cpu)->lock); -#else - data = *cpuusage; #endif return data; @@ -117,69 +145,103 @@ static u64 cpuacct_cpuusage_read(struct cpuacct *ca, int cpu) static void cpuacct_cpuusage_write(struct cpuacct *ca, int cpu, u64 val) { - u64 *cpuusage = per_cpu_ptr(ca->cpuusage, cpu); + struct cpuacct_usage *cpuusage = per_cpu_ptr(ca->cpuusage, cpu); + int i; #ifndef CONFIG_64BIT /* * Take rq->lock to make 64-bit write safe on 32-bit platforms. */ raw_spin_lock_irq(&cpu_rq(cpu)->lock); - *cpuusage = val; +#endif + + for (i = 0; i < CPUACCT_USAGE_NRUSAGE; i++) + cpuusage->usages[i] = val; + +#ifndef CONFIG_64BIT raw_spin_unlock_irq(&cpu_rq(cpu)->lock); -#else - *cpuusage = val; #endif } /* return total cpu usage (in nanoseconds) of a group */ -static u64 cpuusage_read(struct cgroup_subsys_state *css, struct cftype *cft) +static u64 __cpuusage_read(struct cgroup_subsys_state *css, + enum cpuacct_usage_index index) { struct cpuacct *ca = css_ca(css); u64 totalcpuusage = 0; int i; - for_each_present_cpu(i) - totalcpuusage += cpuacct_cpuusage_read(ca, i); + for_each_possible_cpu(i) + totalcpuusage += cpuacct_cpuusage_read(ca, i, index); return totalcpuusage; } +static u64 cpuusage_user_read(struct cgroup_subsys_state *css, + struct cftype *cft) +{ + return __cpuusage_read(css, CPUACCT_USAGE_USER); +} + +static u64 cpuusage_sys_read(struct cgroup_subsys_state *css, + struct cftype *cft) +{ + return __cpuusage_read(css, CPUACCT_USAGE_SYSTEM); +} + +static u64 cpuusage_read(struct cgroup_subsys_state *css, struct cftype *cft) +{ + return __cpuusage_read(css, CPUACCT_USAGE_NRUSAGE); +} + static int cpuusage_write(struct cgroup_subsys_state *css, struct cftype *cft, u64 val) { struct cpuacct *ca = css_ca(css); - int err = 0; - int i; + int cpu; /* * Only allow '0' here to do a reset. */ - if (val) { - err = -EINVAL; - goto out; - } + if (val) + return -EINVAL; - for_each_present_cpu(i) - cpuacct_cpuusage_write(ca, i, 0); + for_each_possible_cpu(cpu) + cpuacct_cpuusage_write(ca, cpu, 0); -out: - return err; + return 0; } -static int cpuacct_percpu_seq_show(struct seq_file *m, void *V) +static int __cpuacct_percpu_seq_show(struct seq_file *m, + enum cpuacct_usage_index index) { struct cpuacct *ca = css_ca(seq_css(m)); u64 percpu; int i; - for_each_present_cpu(i) { - percpu = cpuacct_cpuusage_read(ca, i); + for_each_possible_cpu(i) { + percpu = cpuacct_cpuusage_read(ca, i, index); seq_printf(m, "%llu ", (unsigned long long) percpu); } seq_printf(m, "\n"); return 0; } +static int cpuacct_percpu_user_seq_show(struct seq_file *m, void *V) +{ + return __cpuacct_percpu_seq_show(m, CPUACCT_USAGE_USER); +} + +static int cpuacct_percpu_sys_seq_show(struct seq_file *m, void *V) +{ + return __cpuacct_percpu_seq_show(m, CPUACCT_USAGE_SYSTEM); +} + +static int cpuacct_percpu_seq_show(struct seq_file *m, void *V) +{ + return __cpuacct_percpu_seq_show(m, CPUACCT_USAGE_NRUSAGE); +} + static const char * const cpuacct_stat_desc[] = { [CPUACCT_STAT_USER] = "user", [CPUACCT_STAT_SYSTEM] = "system", @@ -191,7 +253,7 @@ static int cpuacct_stats_show(struct seq_file *sf, void *v) int cpu; s64 val = 0; - for_each_online_cpu(cpu) { + for_each_possible_cpu(cpu) { struct kernel_cpustat *kcpustat = per_cpu_ptr(ca->cpustat, cpu); val += kcpustat->cpustat[CPUTIME_USER]; val += kcpustat->cpustat[CPUTIME_NICE]; @@ -200,7 +262,7 @@ static int cpuacct_stats_show(struct seq_file *sf, void *v) seq_printf(sf, "%s %lld\n", cpuacct_stat_desc[CPUACCT_STAT_USER], val); val = 0; - for_each_online_cpu(cpu) { + for_each_possible_cpu(cpu) { struct kernel_cpustat *kcpustat = per_cpu_ptr(ca->cpustat, cpu); val += kcpustat->cpustat[CPUTIME_SYSTEM]; val += kcpustat->cpustat[CPUTIME_IRQ]; @@ -220,10 +282,26 @@ static struct cftype files[] = { .write_u64 = cpuusage_write, }, { + .name = "usage_user", + .read_u64 = cpuusage_user_read, + }, + { + .name = "usage_sys", + .read_u64 = cpuusage_sys_read, + }, + { .name = "usage_percpu", .seq_show = cpuacct_percpu_seq_show, }, { + .name = "usage_percpu_user", + .seq_show = cpuacct_percpu_user_seq_show, + }, + { + .name = "usage_percpu_sys", + .seq_show = cpuacct_percpu_sys_seq_show, + }, + { .name = "stat", .seq_show = cpuacct_stats_show, }, @@ -238,10 +316,17 @@ static struct cftype files[] = { void cpuacct_charge(struct task_struct *tsk, u64 cputime) { struct cpuacct *ca; + int index = CPUACCT_USAGE_SYSTEM; + struct pt_regs *regs = task_pt_regs(tsk); + + if (regs && user_mode(regs)) + index = CPUACCT_USAGE_USER; rcu_read_lock(); + for (ca = task_ca(tsk); ca; ca = parent_ca(ca)) - *this_cpu_ptr(ca->cpuusage) += cputime; + this_cpu_ptr(ca->cpuusage)->usages[index] += cputime; + rcu_read_unlock(); } diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c index affd97ec9f65..0ac6c84f3371 100644 --- a/kernel/sched/deadline.c +++ b/kernel/sched/deadline.c @@ -591,10 +591,10 @@ static enum hrtimer_restart dl_task_timer(struct hrtimer *timer) struct sched_dl_entity, dl_timer); struct task_struct *p = dl_task_of(dl_se); - unsigned long flags; + struct rq_flags rf; struct rq *rq; - rq = task_rq_lock(p, &flags); + rq = task_rq_lock(p, &rf); /* * The task might have changed its scheduling policy to something @@ -670,14 +670,14 @@ static enum hrtimer_restart dl_task_timer(struct hrtimer *timer) * Nothing relies on rq->lock after this, so its safe to drop * rq->lock. */ - lockdep_unpin_lock(&rq->lock); + lockdep_unpin_lock(&rq->lock, rf.cookie); push_dl_task(rq); - lockdep_pin_lock(&rq->lock); + lockdep_repin_lock(&rq->lock, rf.cookie); } #endif unlock: - task_rq_unlock(rq, p, &flags); + task_rq_unlock(rq, p, &rf); /* * This can free the task_struct, including this hrtimer, do not touch @@ -717,10 +717,6 @@ static void update_curr_dl(struct rq *rq) if (!dl_task(curr) || !on_dl_rq(dl_se)) return; - /* Kick cpufreq (see the comment in linux/cpufreq.h). */ - if (cpu_of(rq) == smp_processor_id()) - cpufreq_trigger_update(rq_clock(rq)); - /* * Consumed budget is computed considering the time as * observed by schedulable tasks (excluding time spent @@ -736,6 +732,10 @@ static void update_curr_dl(struct rq *rq) return; } + /* kick cpufreq (see the comment in linux/cpufreq.h). */ + if (cpu_of(rq) == smp_processor_id()) + cpufreq_trigger_update(rq_clock(rq)); + schedstat_set(curr->se.statistics.exec_max, max(curr->se.statistics.exec_max, delta_exec)); @@ -1125,7 +1125,8 @@ static struct sched_dl_entity *pick_next_dl_entity(struct rq *rq, return rb_entry(left, struct sched_dl_entity, rb_node); } -struct task_struct *pick_next_task_dl(struct rq *rq, struct task_struct *prev) +struct task_struct * +pick_next_task_dl(struct rq *rq, struct task_struct *prev, struct pin_cookie cookie) { struct sched_dl_entity *dl_se; struct task_struct *p; @@ -1140,9 +1141,9 @@ struct task_struct *pick_next_task_dl(struct rq *rq, struct task_struct *prev) * disabled avoiding further scheduler activity on it and we're * being very careful to re-start the picking loop. */ - lockdep_unpin_lock(&rq->lock); + lockdep_unpin_lock(&rq->lock, cookie); pull_dl_task(rq); - lockdep_pin_lock(&rq->lock); + lockdep_repin_lock(&rq->lock, cookie); /* * pull_rt_task() can drop (and re-acquire) rq->lock; this * means a stop task can slip in, in which case we need to @@ -1394,6 +1395,7 @@ static struct rq *find_lock_later_rq(struct task_struct *task, struct rq *rq) !cpumask_test_cpu(later_rq->cpu, &task->cpus_allowed) || task_running(rq, task) || + !dl_task(task) || !task_on_rq_queued(task))) { double_unlock_balance(rq, later_rq); later_rq = NULL; diff --git a/kernel/sched/debug.c b/kernel/sched/debug.c index 4fbc3bd5ff60..cf905f655ba1 100644 --- a/kernel/sched/debug.c +++ b/kernel/sched/debug.c @@ -626,15 +626,16 @@ do { \ #undef P #undef PN -#ifdef CONFIG_SCHEDSTATS -#define P(n) SEQ_printf(m, " .%-30s: %d\n", #n, rq->n); -#define P64(n) SEQ_printf(m, " .%-30s: %Ld\n", #n, rq->n); - #ifdef CONFIG_SMP +#define P64(n) SEQ_printf(m, " .%-30s: %Ld\n", #n, rq->n); P64(avg_idle); P64(max_idle_balance_cost); +#undef P64 #endif +#ifdef CONFIG_SCHEDSTATS +#define P(n) SEQ_printf(m, " .%-30s: %d\n", #n, rq->n); + if (schedstat_enabled()) { P(yld_count); P(sched_count); @@ -644,7 +645,6 @@ do { \ } #undef P -#undef P64 #endif spin_lock_irqsave(&sched_debug_lock, flags); print_cfs_stats(m, cpu); diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 8b6db3626566..84e465ae7c63 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -204,7 +204,7 @@ static void __update_inv_weight(struct load_weight *lw) * OR * (delta_exec * (weight * lw->inv_weight)) >> WMULT_SHIFT * - * Either weight := NICE_0_LOAD and lw \e prio_to_wmult[], in which case + * Either weight := NICE_0_LOAD and lw \e sched_prio_to_wmult[], in which case * we're guaranteed shift stays positive because inv_weight is guaranteed to * fit 32 bits, and NICE_0_LOAD gives another 10 bits; therefore shift >= 22. * @@ -682,17 +682,68 @@ void init_entity_runnable_average(struct sched_entity *se) sa->period_contrib = 1023; sa->load_avg = scale_load_down(se->load.weight); sa->load_sum = sa->load_avg * LOAD_AVG_MAX; - sa->util_avg = scale_load_down(SCHED_LOAD_SCALE); - sa->util_sum = sa->util_avg * LOAD_AVG_MAX; + /* + * At this point, util_avg won't be used in select_task_rq_fair anyway + */ + sa->util_avg = 0; + sa->util_sum = 0; /* when this task enqueue'ed, it will contribute to its cfs_rq's load_avg */ } +/* + * With new tasks being created, their initial util_avgs are extrapolated + * based on the cfs_rq's current util_avg: + * + * util_avg = cfs_rq->util_avg / (cfs_rq->load_avg + 1) * se.load.weight + * + * However, in many cases, the above util_avg does not give a desired + * value. Moreover, the sum of the util_avgs may be divergent, such + * as when the series is a harmonic series. + * + * To solve this problem, we also cap the util_avg of successive tasks to + * only 1/2 of the left utilization budget: + * + * util_avg_cap = (1024 - cfs_rq->avg.util_avg) / 2^n + * + * where n denotes the nth task. + * + * For example, a simplest series from the beginning would be like: + * + * task util_avg: 512, 256, 128, 64, 32, 16, 8, ... + * cfs_rq util_avg: 512, 768, 896, 960, 992, 1008, 1016, ... + * + * Finally, that extrapolated util_avg is clamped to the cap (util_avg_cap) + * if util_avg > util_avg_cap. + */ +void post_init_entity_util_avg(struct sched_entity *se) +{ + struct cfs_rq *cfs_rq = cfs_rq_of(se); + struct sched_avg *sa = &se->avg; + long cap = (long)(SCHED_CAPACITY_SCALE - cfs_rq->avg.util_avg) / 2; + + if (cap > 0) { + if (cfs_rq->avg.util_avg != 0) { + sa->util_avg = cfs_rq->avg.util_avg * se->load.weight; + sa->util_avg /= (cfs_rq->avg.load_avg + 1); + + if (sa->util_avg > cap) + sa->util_avg = cap; + } else { + sa->util_avg = cap; + } + sa->util_sum = sa->util_avg * LOAD_AVG_MAX; + } +} + static inline unsigned long cfs_rq_runnable_load_avg(struct cfs_rq *cfs_rq); static inline unsigned long cfs_rq_load_avg(struct cfs_rq *cfs_rq); #else void init_entity_runnable_average(struct sched_entity *se) { } +void post_init_entity_util_avg(struct sched_entity *se) +{ +} #endif /* @@ -2437,10 +2488,12 @@ account_entity_dequeue(struct cfs_rq *cfs_rq, struct sched_entity *se) update_load_sub(&cfs_rq->load, se->load.weight); if (!parent_entity(se)) update_load_sub(&rq_of(cfs_rq)->load, se->load.weight); +#ifdef CONFIG_SMP if (entity_is_task(se)) { account_numa_dequeue(rq_of(cfs_rq), task_of(se)); list_del_init(&se->group_node); } +#endif cfs_rq->nr_running--; } @@ -2550,6 +2603,16 @@ static const u32 runnable_avg_yN_sum[] = { }; /* + * Precomputed \Sum y^k { 1<=k<=n, where n%32=0). Values are rolled down to + * lower integers. See Documentation/scheduler/sched-avg.txt how these + * were generated: + */ +static const u32 __accumulated_sum_N32[] = { + 0, 23371, 35056, 40899, 43820, 45281, + 46011, 46376, 46559, 46650, 46696, 46719, +}; + +/* * Approximate: * val * y^n, where y^32 ~= 0.5 (~1 scheduling period) */ @@ -2597,22 +2660,13 @@ static u32 __compute_runnable_contrib(u64 n) else if (unlikely(n >= LOAD_AVG_MAX_N)) return LOAD_AVG_MAX; - /* Compute \Sum k^n combining precomputed values for k^i, \Sum k^j */ - do { - contrib /= 2; /* y^LOAD_AVG_PERIOD = 1/2 */ - contrib += runnable_avg_yN_sum[LOAD_AVG_PERIOD]; - - n -= LOAD_AVG_PERIOD; - } while (n > LOAD_AVG_PERIOD); - + /* Since n < LOAD_AVG_MAX_N, n/LOAD_AVG_PERIOD < 11 */ + contrib = __accumulated_sum_N32[n/LOAD_AVG_PERIOD]; + n %= LOAD_AVG_PERIOD; contrib = decay_load(contrib, n); return contrib + runnable_avg_yN_sum[n]; } -#if (SCHED_LOAD_SHIFT - SCHED_LOAD_RESOLUTION) != 10 || SCHED_CAPACITY_SHIFT != 10 -#error "load tracking assumes 2^10 as unit" -#endif - #define cap_scale(v, s) ((v)*(s) >> SCHED_CAPACITY_SHIFT) /* @@ -2821,23 +2875,54 @@ static inline void update_tg_load_avg(struct cfs_rq *cfs_rq, int force) {} static inline u64 cfs_rq_clock_task(struct cfs_rq *cfs_rq); +static inline void cfs_rq_util_change(struct cfs_rq *cfs_rq) +{ + struct rq *rq = rq_of(cfs_rq); + int cpu = cpu_of(rq); + + if (cpu == smp_processor_id() && &rq->cfs == cfs_rq) { + unsigned long max = rq->cpu_capacity_orig; + + /* + * There are a few boundary cases this might miss but it should + * get called often enough that that should (hopefully) not be + * a real problem -- added to that it only calls on the local + * CPU, so if we enqueue remotely we'll miss an update, but + * the next tick/schedule should update. + * + * It will not get called when we go idle, because the idle + * thread is a different class (!fair), nor will the utilization + * number include things like RT tasks. + * + * As is, the util number is not freq-invariant (we'd have to + * implement arch_scale_freq_capacity() for that). + * + * See cpu_util(). + */ + cpufreq_update_util(rq_clock(rq), + min(cfs_rq->avg.util_avg, max), max); + } +} + /* Group cfs_rq's load_avg is used for task_h_load and update_cfs_share */ -static inline int update_cfs_rq_load_avg(u64 now, struct cfs_rq *cfs_rq) +static inline int +update_cfs_rq_load_avg(u64 now, struct cfs_rq *cfs_rq, bool update_freq) { struct sched_avg *sa = &cfs_rq->avg; - int decayed, removed = 0; + int decayed, removed_load = 0, removed_util = 0; if (atomic_long_read(&cfs_rq->removed_load_avg)) { s64 r = atomic_long_xchg(&cfs_rq->removed_load_avg, 0); sa->load_avg = max_t(long, sa->load_avg - r, 0); sa->load_sum = max_t(s64, sa->load_sum - r * LOAD_AVG_MAX, 0); - removed = 1; + removed_load = 1; } if (atomic_long_read(&cfs_rq->removed_util_avg)) { long r = atomic_long_xchg(&cfs_rq->removed_util_avg, 0); sa->util_avg = max_t(long, sa->util_avg - r, 0); sa->util_sum = max_t(s32, sa->util_sum - r * LOAD_AVG_MAX, 0); + removed_util = 1; } decayed = __update_load_avg(now, cpu_of(rq_of(cfs_rq)), sa, @@ -2848,7 +2933,10 @@ static inline int update_cfs_rq_load_avg(u64 now, struct cfs_rq *cfs_rq) cfs_rq->load_last_update_time_copy = sa->last_update_time; #endif - return decayed || removed; + if (update_freq && (decayed || removed_util)) + cfs_rq_util_change(cfs_rq); + + return decayed || removed_load; } /* Update task and its cfs_rq load average */ @@ -2867,31 +2955,8 @@ static inline void update_load_avg(struct sched_entity *se, int update_tg) se->on_rq * scale_load_down(se->load.weight), cfs_rq->curr == se, NULL); - if (update_cfs_rq_load_avg(now, cfs_rq) && update_tg) + if (update_cfs_rq_load_avg(now, cfs_rq, true) && update_tg) update_tg_load_avg(cfs_rq, 0); - - if (cpu == smp_processor_id() && &rq->cfs == cfs_rq) { - unsigned long max = rq->cpu_capacity_orig; - - /* - * There are a few boundary cases this might miss but it should - * get called often enough that that should (hopefully) not be - * a real problem -- added to that it only calls on the local - * CPU, so if we enqueue remotely we'll miss an update, but - * the next tick/schedule should update. - * - * It will not get called when we go idle, because the idle - * thread is a different class (!fair), nor will the utilization - * number include things like RT tasks. - * - * As is, the util number is not freq-invariant (we'd have to - * implement arch_scale_freq_capacity() for that). - * - * See cpu_util(). - */ - cpufreq_update_util(rq_clock(rq), - min(cfs_rq->avg.util_avg, max), max); - } } static void attach_entity_load_avg(struct cfs_rq *cfs_rq, struct sched_entity *se) @@ -2919,6 +2984,8 @@ skip_aging: cfs_rq->avg.load_sum += se->avg.load_sum; cfs_rq->avg.util_avg += se->avg.util_avg; cfs_rq->avg.util_sum += se->avg.util_sum; + + cfs_rq_util_change(cfs_rq); } static void detach_entity_load_avg(struct cfs_rq *cfs_rq, struct sched_entity *se) @@ -2931,6 +2998,8 @@ static void detach_entity_load_avg(struct cfs_rq *cfs_rq, struct sched_entity *s cfs_rq->avg.load_sum = max_t(s64, cfs_rq->avg.load_sum - se->avg.load_sum, 0); cfs_rq->avg.util_avg = max_t(long, cfs_rq->avg.util_avg - se->avg.util_avg, 0); cfs_rq->avg.util_sum = max_t(s32, cfs_rq->avg.util_sum - se->avg.util_sum, 0); + + cfs_rq_util_change(cfs_rq); } /* Add the load generated by se into cfs_rq's load average */ @@ -2948,7 +3017,7 @@ enqueue_entity_load_avg(struct cfs_rq *cfs_rq, struct sched_entity *se) cfs_rq->curr == se, NULL); } - decayed = update_cfs_rq_load_avg(now, cfs_rq); + decayed = update_cfs_rq_load_avg(now, cfs_rq, !migrated); cfs_rq->runnable_load_avg += sa->load_avg; cfs_rq->runnable_load_sum += sa->load_sum; @@ -3030,7 +3099,14 @@ static int idle_balance(struct rq *this_rq); #else /* CONFIG_SMP */ -static inline void update_load_avg(struct sched_entity *se, int update_tg) {} +static inline void update_load_avg(struct sched_entity *se, int not_used) +{ + struct cfs_rq *cfs_rq = cfs_rq_of(se); + struct rq *rq = rq_of(cfs_rq); + + cpufreq_trigger_update(rq_clock(rq)); +} + static inline void enqueue_entity_load_avg(struct cfs_rq *cfs_rq, struct sched_entity *se) {} static inline void @@ -3181,25 +3257,17 @@ static inline void check_schedstat_required(void) static void enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags) { - bool renorm = !(flags & ENQUEUE_WAKEUP) || (flags & ENQUEUE_WAKING); - bool curr = cfs_rq->curr == se; - /* - * If we're the current task, we must renormalise before calling - * update_curr(). + * Update the normalized vruntime before updating min_vruntime + * through calling update_curr(). */ - if (renorm && curr) + if (!(flags & ENQUEUE_WAKEUP) || (flags & ENQUEUE_WAKING)) se->vruntime += cfs_rq->min_vruntime; - update_curr(cfs_rq); - /* - * Otherwise, renormalise after, such that we're placed at the current - * moment in time, instead of some random moment in the past. + * Update run-time statistics of the 'current'. */ - if (renorm && !curr) - se->vruntime += cfs_rq->min_vruntime; - + update_curr(cfs_rq); enqueue_entity_load_avg(cfs_rq, se); account_entity_enqueue(cfs_rq, se); update_cfs_shares(cfs_rq); @@ -3215,7 +3283,7 @@ enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags) update_stats_enqueue(cfs_rq, se); check_spread(cfs_rq, se); } - if (!curr) + if (se != cfs_rq->curr) __enqueue_entity(cfs_rq, se); se->on_rq = 1; @@ -4423,7 +4491,7 @@ static void dequeue_task_fair(struct rq *rq, struct task_struct *p, int flags) } #ifdef CONFIG_SMP - +#ifdef CONFIG_NO_HZ_COMMON /* * per rq 'load' arrray crap; XXX kill this. */ @@ -4489,13 +4557,13 @@ decay_load_missed(unsigned long load, unsigned long missed_updates, int idx) } return load; } +#endif /* CONFIG_NO_HZ_COMMON */ /** - * __update_cpu_load - update the rq->cpu_load[] statistics + * __cpu_load_update - update the rq->cpu_load[] statistics * @this_rq: The rq to update statistics for * @this_load: The current load * @pending_updates: The number of missed updates - * @active: !0 for NOHZ_FULL * * Update rq->cpu_load[] statistics. This function is usually called every * scheduler tick (TICK_NSEC). @@ -4524,12 +4592,12 @@ decay_load_missed(unsigned long load, unsigned long missed_updates, int idx) * load[i]_n = (1 - 1/2^i)^n * load[i]_0 * * see decay_load_misses(). For NOHZ_FULL we get to subtract and add the extra - * term. See the @active paramter. + * term. */ -static void __update_cpu_load(struct rq *this_rq, unsigned long this_load, - unsigned long pending_updates, int active) +static void cpu_load_update(struct rq *this_rq, unsigned long this_load, + unsigned long pending_updates) { - unsigned long tickless_load = active ? this_rq->cpu_load[0] : 0; + unsigned long __maybe_unused tickless_load = this_rq->cpu_load[0]; int i, scale; this_rq->nr_load_updates++; @@ -4542,6 +4610,7 @@ static void __update_cpu_load(struct rq *this_rq, unsigned long this_load, /* scale is effectively 1 << i now, and >> i divides by scale */ old_load = this_rq->cpu_load[i]; +#ifdef CONFIG_NO_HZ_COMMON old_load = decay_load_missed(old_load, pending_updates - 1, i); if (tickless_load) { old_load -= decay_load_missed(tickless_load, pending_updates - 1, i); @@ -4552,6 +4621,7 @@ static void __update_cpu_load(struct rq *this_rq, unsigned long this_load, */ old_load += tickless_load; } +#endif new_load = this_load; /* * Round up the averaging division if load is increasing. This @@ -4574,10 +4644,23 @@ static unsigned long weighted_cpuload(const int cpu) } #ifdef CONFIG_NO_HZ_COMMON -static void __update_cpu_load_nohz(struct rq *this_rq, - unsigned long curr_jiffies, - unsigned long load, - int active) +/* + * There is no sane way to deal with nohz on smp when using jiffies because the + * cpu doing the jiffies update might drift wrt the cpu doing the jiffy reading + * causing off-by-one errors in observed deltas; {0,2} instead of {1,1}. + * + * Therefore we need to avoid the delta approach from the regular tick when + * possible since that would seriously skew the load calculation. This is why we + * use cpu_load_update_periodic() for CPUs out of nohz. However we'll rely on + * jiffies deltas for updates happening while in nohz mode (idle ticks, idle + * loop exit, nohz_idle_balance, nohz full exit...) + * + * This means we might still be one tick off for nohz periods. + */ + +static void cpu_load_update_nohz(struct rq *this_rq, + unsigned long curr_jiffies, + unsigned long load) { unsigned long pending_updates; @@ -4589,28 +4672,15 @@ static void __update_cpu_load_nohz(struct rq *this_rq, * In the NOHZ_FULL case, we were non-idle, we should consider * its weighted load. */ - __update_cpu_load(this_rq, load, pending_updates, active); + cpu_load_update(this_rq, load, pending_updates); } } /* - * There is no sane way to deal with nohz on smp when using jiffies because the - * cpu doing the jiffies update might drift wrt the cpu doing the jiffy reading - * causing off-by-one errors in observed deltas; {0,2} instead of {1,1}. - * - * Therefore we cannot use the delta approach from the regular tick since that - * would seriously skew the load calculation. However we'll make do for those - * updates happening while idle (nohz_idle_balance) or coming out of idle - * (tick_nohz_idle_exit). - * - * This means we might still be one tick off for nohz periods. - */ - -/* * Called from nohz_idle_balance() to update the load ratings before doing the * idle balance. */ -static void update_cpu_load_idle(struct rq *this_rq) +static void cpu_load_update_idle(struct rq *this_rq) { /* * bail if there's load or we're actually up-to-date. @@ -4618,38 +4688,71 @@ static void update_cpu_load_idle(struct rq *this_rq) if (weighted_cpuload(cpu_of(this_rq))) return; - __update_cpu_load_nohz(this_rq, READ_ONCE(jiffies), 0, 0); + cpu_load_update_nohz(this_rq, READ_ONCE(jiffies), 0); } /* - * Called from tick_nohz_idle_exit() -- try and fix up the ticks we missed. + * Record CPU load on nohz entry so we know the tickless load to account + * on nohz exit. cpu_load[0] happens then to be updated more frequently + * than other cpu_load[idx] but it should be fine as cpu_load readers + * shouldn't rely into synchronized cpu_load[*] updates. */ -void update_cpu_load_nohz(int active) +void cpu_load_update_nohz_start(void) { struct rq *this_rq = this_rq(); + + /* + * This is all lockless but should be fine. If weighted_cpuload changes + * concurrently we'll exit nohz. And cpu_load write can race with + * cpu_load_update_idle() but both updater would be writing the same. + */ + this_rq->cpu_load[0] = weighted_cpuload(cpu_of(this_rq)); +} + +/* + * Account the tickless load in the end of a nohz frame. + */ +void cpu_load_update_nohz_stop(void) +{ unsigned long curr_jiffies = READ_ONCE(jiffies); - unsigned long load = active ? weighted_cpuload(cpu_of(this_rq)) : 0; + struct rq *this_rq = this_rq(); + unsigned long load; if (curr_jiffies == this_rq->last_load_update_tick) return; + load = weighted_cpuload(cpu_of(this_rq)); raw_spin_lock(&this_rq->lock); - __update_cpu_load_nohz(this_rq, curr_jiffies, load, active); + update_rq_clock(this_rq); + cpu_load_update_nohz(this_rq, curr_jiffies, load); raw_spin_unlock(&this_rq->lock); } -#endif /* CONFIG_NO_HZ */ +#else /* !CONFIG_NO_HZ_COMMON */ +static inline void cpu_load_update_nohz(struct rq *this_rq, + unsigned long curr_jiffies, + unsigned long load) { } +#endif /* CONFIG_NO_HZ_COMMON */ + +static void cpu_load_update_periodic(struct rq *this_rq, unsigned long load) +{ +#ifdef CONFIG_NO_HZ_COMMON + /* See the mess around cpu_load_update_nohz(). */ + this_rq->last_load_update_tick = READ_ONCE(jiffies); +#endif + cpu_load_update(this_rq, load, 1); +} /* * Called from scheduler_tick() */ -void update_cpu_load_active(struct rq *this_rq) +void cpu_load_update_active(struct rq *this_rq) { unsigned long load = weighted_cpuload(cpu_of(this_rq)); - /* - * See the mess around update_cpu_load_idle() / update_cpu_load_nohz(). - */ - this_rq->last_load_update_tick = jiffies; - __update_cpu_load(this_rq, load, 1, 1); + + if (tick_nohz_tick_stopped()) + cpu_load_update_nohz(this_rq, READ_ONCE(jiffies), load); + else + cpu_load_update_periodic(this_rq, load); } /* @@ -5440,7 +5543,7 @@ preempt: } static struct task_struct * -pick_next_task_fair(struct rq *rq, struct task_struct *prev) +pick_next_task_fair(struct rq *rq, struct task_struct *prev, struct pin_cookie cookie) { struct cfs_rq *cfs_rq = &rq->cfs; struct sched_entity *se; @@ -5553,9 +5656,9 @@ idle: * further scheduler activity on it and we're being very careful to * re-start the picking loop. */ - lockdep_unpin_lock(&rq->lock); + lockdep_unpin_lock(&rq->lock, cookie); new_tasks = idle_balance(rq); - lockdep_pin_lock(&rq->lock); + lockdep_repin_lock(&rq->lock, cookie); /* * Because idle_balance() releases (and re-acquires) rq->lock, it is * possible for any higher priority task to appear. In that case we @@ -5654,7 +5757,7 @@ static bool yield_to_task_fair(struct rq *rq, struct task_struct *p, bool preemp * W_i,0 = \Sum_j w_i,j (2) * * Where w_i,j is the weight of the j-th runnable task on cpu i. This weight - * is derived from the nice value as per prio_to_weight[]. + * is derived from the nice value as per sched_prio_to_weight[]. * * The weight average is an exponential decay average of the instantaneous * weight: @@ -6156,7 +6259,7 @@ static void update_blocked_averages(int cpu) if (throttled_hierarchy(cfs_rq)) continue; - if (update_cfs_rq_load_avg(cfs_rq_clock_task(cfs_rq), cfs_rq)) + if (update_cfs_rq_load_avg(cfs_rq_clock_task(cfs_rq), cfs_rq, true)) update_tg_load_avg(cfs_rq, 0); } raw_spin_unlock_irqrestore(&rq->lock, flags); @@ -6217,7 +6320,7 @@ static inline void update_blocked_averages(int cpu) raw_spin_lock_irqsave(&rq->lock, flags); update_rq_clock(rq); - update_cfs_rq_load_avg(cfs_rq_clock_task(cfs_rq), cfs_rq); + update_cfs_rq_load_avg(cfs_rq_clock_task(cfs_rq), cfs_rq, true); raw_spin_unlock_irqrestore(&rq->lock, flags); } @@ -6626,6 +6729,9 @@ static bool update_sd_pick_busiest(struct lb_env *env, if (!(env->sd->flags & SD_ASYM_PACKING)) return true; + /* No ASYM_PACKING if target cpu is already busy */ + if (env->idle == CPU_NOT_IDLE) + return true; /* * ASYM_PACKING needs to move all the work to the lowest * numbered CPUs in the group, therefore mark all groups @@ -6635,7 +6741,8 @@ static bool update_sd_pick_busiest(struct lb_env *env, if (!sds->busiest) return true; - if (group_first_cpu(sds->busiest) > group_first_cpu(sg)) + /* Prefer to move from highest possible cpu's work */ + if (group_first_cpu(sds->busiest) < group_first_cpu(sg)) return true; } @@ -6781,6 +6888,9 @@ static int check_asym_packing(struct lb_env *env, struct sd_lb_stats *sds) if (!(env->sd->flags & SD_ASYM_PACKING)) return 0; + if (env->idle == CPU_NOT_IDLE) + return 0; + if (!sds->busiest) return 0; @@ -6889,9 +6999,10 @@ static inline void calculate_imbalance(struct lb_env *env, struct sd_lb_stats *s } /* - * In the presence of smp nice balancing, certain scenarios can have - * max load less than avg load(as we skip the groups at or below - * its cpu_capacity, while calculating max_load..) + * Avg load of busiest sg can be less and avg load of local sg can + * be greater than avg load across all sgs of sd because avg load + * factors in sg capacity and sgs with smaller group_type are + * skipped when updating the busiest sg: */ if (busiest->avg_load <= sds->avg_load || local->avg_load >= sds->avg_load) { @@ -6905,7 +7016,7 @@ static inline void calculate_imbalance(struct lb_env *env, struct sd_lb_stats *s if (busiest->group_type == group_overloaded && local->group_type == group_overloaded) { load_above_capacity = busiest->sum_nr_running * - SCHED_LOAD_SCALE; + scale_load_down(NICE_0_LOAD); if (load_above_capacity > busiest->group_capacity) load_above_capacity -= busiest->group_capacity; else @@ -6916,9 +7027,8 @@ static inline void calculate_imbalance(struct lb_env *env, struct sd_lb_stats *s * We're trying to get all the cpus to the average_load, so we don't * want to push ourselves above the average load, nor do we wish to * reduce the max loaded cpu below the average load. At the same time, - * we also don't want to reduce the group load below the group capacity - * (so that we can implement power-savings policies etc). Thus we look - * for the minimum possible imbalance. + * we also don't want to reduce the group load below the group + * capacity. Thus we look for the minimum possible imbalance. */ max_pull = min(busiest->avg_load - sds->avg_load, load_above_capacity); @@ -6942,10 +7052,7 @@ static inline void calculate_imbalance(struct lb_env *env, struct sd_lb_stats *s /** * find_busiest_group - Returns the busiest group within the sched_domain - * if there is an imbalance. If there isn't an imbalance, and - * the user has opted for power-savings, it returns a group whose - * CPUs can be put to idle by rebalancing those tasks elsewhere, if - * such a group exists. + * if there is an imbalance. * * Also calculates the amount of weighted load which should be moved * to restore balance. @@ -6953,9 +7060,6 @@ static inline void calculate_imbalance(struct lb_env *env, struct sd_lb_stats *s * @env: The load balancing environment. * * Return: - The busiest group if imbalance exists. - * - If no imbalance and user has opted for power-savings balance, - * return the least loaded group whose CPUs can be - * put to idle by rebalancing its tasks onto our group. */ static struct sched_group *find_busiest_group(struct lb_env *env) { @@ -6973,8 +7077,7 @@ static struct sched_group *find_busiest_group(struct lb_env *env) busiest = &sds.busiest_stat; /* ASYM feature bypasses nice load balance check */ - if ((env->idle == CPU_IDLE || env->idle == CPU_NEWLY_IDLE) && - check_asym_packing(env, &sds)) + if (check_asym_packing(env, &sds)) return sds.busiest; /* There is no busy sibling group to pull tasks from */ @@ -7399,10 +7502,7 @@ more_balance: &busiest->active_balance_work); } - /* - * We've kicked active balancing, reset the failure - * counter. - */ + /* We've kicked active balancing, force task migration. */ sd->nr_balance_failed = sd->cache_nice_tries+1; } } else @@ -7637,10 +7737,13 @@ static int active_load_balance_cpu_stop(void *data) schedstat_inc(sd, alb_count); p = detach_one_task(&env); - if (p) + if (p) { schedstat_inc(sd, alb_pushed); - else + /* Active balancing done, reset the failure counter. */ + sd->nr_balance_failed = 0; + } else { schedstat_inc(sd, alb_failed); + } } rcu_read_unlock(); out_unlock: @@ -7945,7 +8048,7 @@ static void nohz_idle_balance(struct rq *this_rq, enum cpu_idle_type idle) if (time_after_eq(jiffies, rq->next_balance)) { raw_spin_lock_irq(&rq->lock); update_rq_clock(rq); - update_cpu_load_idle(rq); + cpu_load_update_idle(rq); raw_spin_unlock_irq(&rq->lock); rebalance_domains(rq, CPU_IDLE); } @@ -8370,6 +8473,7 @@ int alloc_fair_sched_group(struct task_group *tg, struct task_group *parent) init_cfs_rq(cfs_rq); init_tg_cfs_entry(tg, cfs_rq, se, i, parent->se[i]); init_entity_runnable_average(se); + post_init_entity_util_avg(se); } return 1; diff --git a/kernel/sched/idle_task.c b/kernel/sched/idle_task.c index 47ce94931f1b..2ce5458bbe1d 100644 --- a/kernel/sched/idle_task.c +++ b/kernel/sched/idle_task.c @@ -24,7 +24,7 @@ static void check_preempt_curr_idle(struct rq *rq, struct task_struct *p, int fl } static struct task_struct * -pick_next_task_idle(struct rq *rq, struct task_struct *prev) +pick_next_task_idle(struct rq *rq, struct task_struct *prev, struct pin_cookie cookie) { put_prev_task(rq, prev); diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c index c41ea7ac1764..67afa06cc8bc 100644 --- a/kernel/sched/rt.c +++ b/kernel/sched/rt.c @@ -953,14 +953,14 @@ static void update_curr_rt(struct rq *rq) if (curr->sched_class != &rt_sched_class) return; - /* Kick cpufreq (see the comment in linux/cpufreq.h). */ - if (cpu_of(rq) == smp_processor_id()) - cpufreq_trigger_update(rq_clock(rq)); - delta_exec = rq_clock_task(rq) - curr->se.exec_start; if (unlikely((s64)delta_exec <= 0)) return; + /* Kick cpufreq (see the comment in linux/cpufreq.h). */ + if (cpu_of(rq) == smp_processor_id()) + cpufreq_trigger_update(rq_clock(rq)); + schedstat_set(curr->se.statistics.exec_max, max(curr->se.statistics.exec_max, delta_exec)); @@ -1524,7 +1524,7 @@ static struct task_struct *_pick_next_task_rt(struct rq *rq) } static struct task_struct * -pick_next_task_rt(struct rq *rq, struct task_struct *prev) +pick_next_task_rt(struct rq *rq, struct task_struct *prev, struct pin_cookie cookie) { struct task_struct *p; struct rt_rq *rt_rq = &rq->rt; @@ -1536,9 +1536,9 @@ pick_next_task_rt(struct rq *rq, struct task_struct *prev) * disabled avoiding further scheduler activity on it and we're * being very careful to re-start the picking loop. */ - lockdep_unpin_lock(&rq->lock); + lockdep_unpin_lock(&rq->lock, cookie); pull_rt_task(rq); - lockdep_pin_lock(&rq->lock); + lockdep_repin_lock(&rq->lock, cookie); /* * pull_rt_task() can drop (and re-acquire) rq->lock; this * means a dl or stop task can slip in, in which case we need @@ -1729,6 +1729,7 @@ static struct rq *find_lock_lowest_rq(struct task_struct *task, struct rq *rq) !cpumask_test_cpu(lowest_rq->cpu, tsk_cpus_allowed(task)) || task_running(rq, task) || + !rt_task(task) || !task_on_rq_queued(task))) { double_unlock_balance(rq, lowest_rq); diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h index 16a27b624ee5..ab6adb159e23 100644 --- a/kernel/sched/sched.h +++ b/kernel/sched/sched.h @@ -31,9 +31,9 @@ extern void calc_global_load_tick(struct rq *this_rq); extern long calc_load_fold_active(struct rq *this_rq); #ifdef CONFIG_SMP -extern void update_cpu_load_active(struct rq *this_rq); +extern void cpu_load_update_active(struct rq *this_rq); #else -static inline void update_cpu_load_active(struct rq *this_rq) { } +static inline void cpu_load_update_active(struct rq *this_rq) { } #endif /* @@ -49,25 +49,32 @@ static inline void update_cpu_load_active(struct rq *this_rq) { } * and does not change the user-interface for setting shares/weights. * * We increase resolution only if we have enough bits to allow this increased - * resolution (i.e. BITS_PER_LONG > 32). The costs for increasing resolution - * when BITS_PER_LONG <= 32 are pretty high and the returns do not justify the - * increased costs. + * resolution (i.e. 64bit). The costs for increasing resolution when 32bit are + * pretty high and the returns do not justify the increased costs. + * + * Really only required when CONFIG_FAIR_GROUP_SCHED is also set, but to + * increase coverage and consistency always enable it on 64bit platforms. */ -#if 0 /* BITS_PER_LONG > 32 -- currently broken: it increases power usage under light load */ -# define SCHED_LOAD_RESOLUTION 10 -# define scale_load(w) ((w) << SCHED_LOAD_RESOLUTION) -# define scale_load_down(w) ((w) >> SCHED_LOAD_RESOLUTION) +#ifdef CONFIG_64BIT +# define NICE_0_LOAD_SHIFT (SCHED_FIXEDPOINT_SHIFT + SCHED_FIXEDPOINT_SHIFT) +# define scale_load(w) ((w) << SCHED_FIXEDPOINT_SHIFT) +# define scale_load_down(w) ((w) >> SCHED_FIXEDPOINT_SHIFT) #else -# define SCHED_LOAD_RESOLUTION 0 +# define NICE_0_LOAD_SHIFT (SCHED_FIXEDPOINT_SHIFT) # define scale_load(w) (w) # define scale_load_down(w) (w) #endif -#define SCHED_LOAD_SHIFT (10 + SCHED_LOAD_RESOLUTION) -#define SCHED_LOAD_SCALE (1L << SCHED_LOAD_SHIFT) - -#define NICE_0_LOAD SCHED_LOAD_SCALE -#define NICE_0_SHIFT SCHED_LOAD_SHIFT +/* + * Task weight (visible to users) and its load (invisible to users) have + * independent resolution, but they should be well calibrated. We use + * scale_load() and scale_load_down(w) to convert between them. The + * following must be true: + * + * scale_load(sched_prio_to_weight[USER_PRIO(NICE_TO_PRIO(0))]) == NICE_0_LOAD + * + */ +#define NICE_0_LOAD (1L << NICE_0_LOAD_SHIFT) /* * Single value that decides SCHED_DEADLINE internal math precision. @@ -585,11 +592,13 @@ struct rq { #endif #define CPU_LOAD_IDX_MAX 5 unsigned long cpu_load[CPU_LOAD_IDX_MAX]; - unsigned long last_load_update_tick; #ifdef CONFIG_NO_HZ_COMMON +#ifdef CONFIG_SMP + unsigned long last_load_update_tick; +#endif /* CONFIG_SMP */ u64 nohz_stamp; unsigned long nohz_flags; -#endif +#endif /* CONFIG_NO_HZ_COMMON */ #ifdef CONFIG_NO_HZ_FULL unsigned long last_sched_tick; #endif @@ -854,7 +863,7 @@ DECLARE_PER_CPU(struct sched_domain *, sd_asym); struct sched_group_capacity { atomic_t ref; /* - * CPU capacity of this group, SCHED_LOAD_SCALE being max capacity + * CPU capacity of this group, SCHED_CAPACITY_SCALE being max capacity * for a single CPU. */ unsigned int capacity; @@ -1200,7 +1209,8 @@ struct sched_class { * tasks. */ struct task_struct * (*pick_next_task) (struct rq *rq, - struct task_struct *prev); + struct task_struct *prev, + struct pin_cookie cookie); void (*put_prev_task) (struct rq *rq, struct task_struct *p); #ifdef CONFIG_SMP @@ -1313,6 +1323,7 @@ extern void init_dl_task_timer(struct sched_dl_entity *dl_se); unsigned long to_ratio(u64 period, u64 runtime); extern void init_entity_runnable_average(struct sched_entity *se); +extern void post_init_entity_util_avg(struct sched_entity *se); #ifdef CONFIG_NO_HZ_FULL extern bool sched_can_stop_tick(struct rq *rq); @@ -1448,86 +1459,32 @@ static inline void sched_rt_avg_update(struct rq *rq, u64 rt_delta) { } static inline void sched_avg_update(struct rq *rq) { } #endif -/* - * __task_rq_lock - lock the rq @p resides on. - */ -static inline struct rq *__task_rq_lock(struct task_struct *p) - __acquires(rq->lock) -{ - struct rq *rq; - - lockdep_assert_held(&p->pi_lock); - - for (;;) { - rq = task_rq(p); - raw_spin_lock(&rq->lock); - if (likely(rq == task_rq(p) && !task_on_rq_migrating(p))) { - lockdep_pin_lock(&rq->lock); - return rq; - } - raw_spin_unlock(&rq->lock); - - while (unlikely(task_on_rq_migrating(p))) - cpu_relax(); - } -} +struct rq_flags { + unsigned long flags; + struct pin_cookie cookie; +}; -/* - * task_rq_lock - lock p->pi_lock and lock the rq @p resides on. - */ -static inline struct rq *task_rq_lock(struct task_struct *p, unsigned long *flags) +struct rq *__task_rq_lock(struct task_struct *p, struct rq_flags *rf) + __acquires(rq->lock); +struct rq *task_rq_lock(struct task_struct *p, struct rq_flags *rf) __acquires(p->pi_lock) - __acquires(rq->lock) -{ - struct rq *rq; - - for (;;) { - raw_spin_lock_irqsave(&p->pi_lock, *flags); - rq = task_rq(p); - raw_spin_lock(&rq->lock); - /* - * move_queued_task() task_rq_lock() - * - * ACQUIRE (rq->lock) - * [S] ->on_rq = MIGRATING [L] rq = task_rq() - * WMB (__set_task_cpu()) ACQUIRE (rq->lock); - * [S] ->cpu = new_cpu [L] task_rq() - * [L] ->on_rq - * RELEASE (rq->lock) - * - * If we observe the old cpu in task_rq_lock, the acquire of - * the old rq->lock will fully serialize against the stores. - * - * If we observe the new cpu in task_rq_lock, the acquire will - * pair with the WMB to ensure we must then also see migrating. - */ - if (likely(rq == task_rq(p) && !task_on_rq_migrating(p))) { - lockdep_pin_lock(&rq->lock); - return rq; - } - raw_spin_unlock(&rq->lock); - raw_spin_unlock_irqrestore(&p->pi_lock, *flags); - - while (unlikely(task_on_rq_migrating(p))) - cpu_relax(); - } -} + __acquires(rq->lock); -static inline void __task_rq_unlock(struct rq *rq) +static inline void __task_rq_unlock(struct rq *rq, struct rq_flags *rf) __releases(rq->lock) { - lockdep_unpin_lock(&rq->lock); + lockdep_unpin_lock(&rq->lock, rf->cookie); raw_spin_unlock(&rq->lock); } static inline void -task_rq_unlock(struct rq *rq, struct task_struct *p, unsigned long *flags) +task_rq_unlock(struct rq *rq, struct task_struct *p, struct rq_flags *rf) __releases(rq->lock) __releases(p->pi_lock) { - lockdep_unpin_lock(&rq->lock); + lockdep_unpin_lock(&rq->lock, rf->cookie); raw_spin_unlock(&rq->lock); - raw_spin_unlock_irqrestore(&p->pi_lock, *flags); + raw_spin_unlock_irqrestore(&p->pi_lock, rf->flags); } #ifdef CONFIG_SMP diff --git a/kernel/sched/stop_task.c b/kernel/sched/stop_task.c index cbc67da10954..604297a08b3a 100644 --- a/kernel/sched/stop_task.c +++ b/kernel/sched/stop_task.c @@ -24,7 +24,7 @@ check_preempt_curr_stop(struct rq *rq, struct task_struct *p, int flags) } static struct task_struct * -pick_next_task_stop(struct rq *rq, struct task_struct *prev) +pick_next_task_stop(struct rq *rq, struct task_struct *prev, struct pin_cookie cookie) { struct task_struct *stop = rq->stop; |
