diff options
Diffstat (limited to 'kernel/sched_fair.c')
| -rw-r--r-- | kernel/sched_fair.c | 906 |
1 files changed, 537 insertions, 369 deletions
diff --git a/kernel/sched_fair.c b/kernel/sched_fair.c index 5a5ea2cd924..806d1b227a2 100644 --- a/kernel/sched_fair.c +++ b/kernel/sched_fair.c @@ -35,8 +35,8 @@ * (to see the precise effective timeslice length of your workload, * run vmstat and monitor the context-switches (cs) field) */ -unsigned int sysctl_sched_latency = 5000000ULL; -unsigned int normalized_sysctl_sched_latency = 5000000ULL; +unsigned int sysctl_sched_latency = 6000000ULL; +unsigned int normalized_sysctl_sched_latency = 6000000ULL; /* * The initial- and re-scaling of tunables is configurable @@ -52,15 +52,15 @@ enum sched_tunable_scaling sysctl_sched_tunable_scaling /* * Minimal preemption granularity for CPU-bound tasks: - * (default: 1 msec * (1 + ilog(ncpus)), units: nanoseconds) + * (default: 2 msec * (1 + ilog(ncpus)), units: nanoseconds) */ -unsigned int sysctl_sched_min_granularity = 1000000ULL; -unsigned int normalized_sysctl_sched_min_granularity = 1000000ULL; +unsigned int sysctl_sched_min_granularity = 2000000ULL; +unsigned int normalized_sysctl_sched_min_granularity = 2000000ULL; /* * is kept at sysctl_sched_latency / sysctl_sched_min_granularity */ -static unsigned int sched_nr_latency = 5; +static unsigned int sched_nr_latency = 3; /* * After fork, child runs first. If set to 0 (default) then @@ -505,7 +505,8 @@ __update_curr(struct cfs_rq *cfs_rq, struct sched_entity *curr, { unsigned long delta_exec_weighted; - schedstat_set(curr->exec_max, max((u64)delta_exec, curr->exec_max)); + schedstat_set(curr->statistics.exec_max, + max((u64)delta_exec, curr->statistics.exec_max)); curr->sum_exec_runtime += delta_exec; schedstat_add(cfs_rq, exec_clock, delta_exec); @@ -548,7 +549,7 @@ static void update_curr(struct cfs_rq *cfs_rq) static inline void update_stats_wait_start(struct cfs_rq *cfs_rq, struct sched_entity *se) { - schedstat_set(se->wait_start, rq_of(cfs_rq)->clock); + schedstat_set(se->statistics.wait_start, rq_of(cfs_rq)->clock); } /* @@ -567,18 +568,18 @@ static void update_stats_enqueue(struct cfs_rq *cfs_rq, struct sched_entity *se) static void update_stats_wait_end(struct cfs_rq *cfs_rq, struct sched_entity *se) { - schedstat_set(se->wait_max, max(se->wait_max, - rq_of(cfs_rq)->clock - se->wait_start)); - schedstat_set(se->wait_count, se->wait_count + 1); - schedstat_set(se->wait_sum, se->wait_sum + - rq_of(cfs_rq)->clock - se->wait_start); + schedstat_set(se->statistics.wait_max, max(se->statistics.wait_max, + rq_of(cfs_rq)->clock - se->statistics.wait_start)); + schedstat_set(se->statistics.wait_count, se->statistics.wait_count + 1); + schedstat_set(se->statistics.wait_sum, se->statistics.wait_sum + + rq_of(cfs_rq)->clock - se->statistics.wait_start); #ifdef CONFIG_SCHEDSTATS if (entity_is_task(se)) { trace_sched_stat_wait(task_of(se), - rq_of(cfs_rq)->clock - se->wait_start); + rq_of(cfs_rq)->clock - se->statistics.wait_start); } #endif - schedstat_set(se->wait_start, 0); + schedstat_set(se->statistics.wait_start, 0); } static inline void @@ -657,39 +658,39 @@ static void enqueue_sleeper(struct cfs_rq *cfs_rq, struct sched_entity *se) if (entity_is_task(se)) tsk = task_of(se); - if (se->sleep_start) { - u64 delta = rq_of(cfs_rq)->clock - se->sleep_start; + if (se->statistics.sleep_start) { + u64 delta = rq_of(cfs_rq)->clock - se->statistics.sleep_start; if ((s64)delta < 0) delta = 0; - if (unlikely(delta > se->sleep_max)) - se->sleep_max = delta; + if (unlikely(delta > se->statistics.sleep_max)) + se->statistics.sleep_max = delta; - se->sleep_start = 0; - se->sum_sleep_runtime += delta; + se->statistics.sleep_start = 0; + se->statistics.sum_sleep_runtime += delta; if (tsk) { account_scheduler_latency(tsk, delta >> 10, 1); trace_sched_stat_sleep(tsk, delta); } } - if (se->block_start) { - u64 delta = rq_of(cfs_rq)->clock - se->block_start; + if (se->statistics.block_start) { + u64 delta = rq_of(cfs_rq)->clock - se->statistics.block_start; if ((s64)delta < 0) delta = 0; - if (unlikely(delta > se->block_max)) - se->block_max = delta; + if (unlikely(delta > se->statistics.block_max)) + se->statistics.block_max = delta; - se->block_start = 0; - se->sum_sleep_runtime += delta; + se->statistics.block_start = 0; + se->statistics.sum_sleep_runtime += delta; if (tsk) { if (tsk->in_iowait) { - se->iowait_sum += delta; - se->iowait_count++; + se->statistics.iowait_sum += delta; + se->statistics.iowait_count++; trace_sched_stat_iowait(tsk, delta); } @@ -737,20 +738,10 @@ place_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int initial) vruntime += sched_vslice(cfs_rq, se); /* sleeps up to a single latency don't count. */ - if (!initial && sched_feat(FAIR_SLEEPERS)) { + if (!initial) { unsigned long thresh = sysctl_sched_latency; /* - * Convert the sleeper threshold into virtual time. - * SCHED_IDLE is a special sub-class. We care about - * fairness only relative to other SCHED_IDLE tasks, - * all of which have the same weight. - */ - if (sched_feat(NORMALIZED_SLEEPER) && (!entity_is_task(se) || - task_of(se)->policy != SCHED_IDLE)) - thresh = calc_delta_fair(thresh, se); - - /* * Halve their sleep time's effect, to allow * for a gentler effect of sleepers: */ @@ -766,9 +757,6 @@ place_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int initial) se->vruntime = vruntime; } -#define ENQUEUE_WAKEUP 1 -#define ENQUEUE_MIGRATE 2 - static void enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags) { @@ -776,7 +764,7 @@ enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags) * Update the normalized vruntime before updating min_vruntime * through callig update_curr(). */ - if (!(flags & ENQUEUE_WAKEUP) || (flags & ENQUEUE_MIGRATE)) + if (!(flags & ENQUEUE_WAKEUP) || (flags & ENQUEUE_WAKING)) se->vruntime += cfs_rq->min_vruntime; /* @@ -812,7 +800,7 @@ static void clear_buddies(struct cfs_rq *cfs_rq, struct sched_entity *se) } static void -dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int sleep) +dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags) { /* * Update run-time statistics of the 'current'. @@ -820,15 +808,15 @@ dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int sleep) update_curr(cfs_rq); update_stats_dequeue(cfs_rq, se); - if (sleep) { + if (flags & DEQUEUE_SLEEP) { #ifdef CONFIG_SCHEDSTATS if (entity_is_task(se)) { struct task_struct *tsk = task_of(se); if (tsk->state & TASK_INTERRUPTIBLE) - se->sleep_start = rq_of(cfs_rq)->clock; + se->statistics.sleep_start = rq_of(cfs_rq)->clock; if (tsk->state & TASK_UNINTERRUPTIBLE) - se->block_start = rq_of(cfs_rq)->clock; + se->statistics.block_start = rq_of(cfs_rq)->clock; } #endif } @@ -845,7 +833,7 @@ dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int sleep) * update can refer to the ->curr item and we need to reflect this * movement in our normalized position. */ - if (!sleep) + if (!(flags & DEQUEUE_SLEEP)) se->vruntime -= cfs_rq->min_vruntime; } @@ -912,7 +900,7 @@ set_next_entity(struct cfs_rq *cfs_rq, struct sched_entity *se) * when there are only lesser-weight tasks around): */ if (rq_of(cfs_rq)->load.weight >= 2*se->load.weight) { - se->slice_max = max(se->slice_max, + se->statistics.slice_max = max(se->statistics.slice_max, se->sum_exec_runtime - se->prev_sum_exec_runtime); } #endif @@ -1054,16 +1042,10 @@ static inline void hrtick_update(struct rq *rq) * then put the task into the rbtree: */ static void -enqueue_task_fair(struct rq *rq, struct task_struct *p, int wakeup, bool head) +enqueue_task_fair(struct rq *rq, struct task_struct *p, int flags) { struct cfs_rq *cfs_rq; struct sched_entity *se = &p->se; - int flags = 0; - - if (wakeup) - flags |= ENQUEUE_WAKEUP; - if (p->state == TASK_WAKING) - flags |= ENQUEUE_MIGRATE; for_each_sched_entity(se) { if (se->on_rq) @@ -1081,18 +1063,18 @@ enqueue_task_fair(struct rq *rq, struct task_struct *p, int wakeup, bool head) * decreased. We remove the task from the rbtree and * update the fair scheduling stats: */ -static void dequeue_task_fair(struct rq *rq, struct task_struct *p, int sleep) +static void dequeue_task_fair(struct rq *rq, struct task_struct *p, int flags) { struct cfs_rq *cfs_rq; struct sched_entity *se = &p->se; for_each_sched_entity(se) { cfs_rq = cfs_rq_of(se); - dequeue_entity(cfs_rq, se, sleep); + dequeue_entity(cfs_rq, se, flags); /* Don't dequeue parent if it has other entities besides us */ if (cfs_rq->load.weight) break; - sleep = 1; + flags |= DEQUEUE_SLEEP; } hrtick_update(rq); @@ -1240,11 +1222,9 @@ static inline unsigned long effective_load(struct task_group *tg, int cpu, static int wake_affine(struct sched_domain *sd, struct task_struct *p, int sync) { - struct task_struct *curr = current; unsigned long this_load, load; int idx, this_cpu, prev_cpu; unsigned long tl_per_task; - unsigned int imbalance; struct task_group *tg; unsigned long weight; int balanced; @@ -1255,23 +1235,12 @@ static int wake_affine(struct sched_domain *sd, struct task_struct *p, int sync) load = source_load(prev_cpu, idx); this_load = target_load(this_cpu, idx); - if (sync) { - if (sched_feat(SYNC_LESS) && - (curr->se.avg_overlap > sysctl_sched_migration_cost || - p->se.avg_overlap > sysctl_sched_migration_cost)) - sync = 0; - } else { - if (sched_feat(SYNC_MORE) && - (curr->se.avg_overlap < sysctl_sched_migration_cost && - p->se.avg_overlap < sysctl_sched_migration_cost)) - sync = 1; - } - /* * If sync wakeup then subtract the (maximum possible) * effect of the currently running task from the load * of the current CPU: */ + rcu_read_lock(); if (sync) { tg = task_group(current); weight = current->se.load.weight; @@ -1283,8 +1252,6 @@ static int wake_affine(struct sched_domain *sd, struct task_struct *p, int sync) tg = task_group(p); weight = p->se.load.weight; - imbalance = 100 + (sd->imbalance_pct - 100) / 2; - /* * In low-load situations, where prev_cpu is idle and this_cpu is idle * due to the sync cause above having dropped this_load to 0, we'll @@ -1294,9 +1261,22 @@ static int wake_affine(struct sched_domain *sd, struct task_struct *p, int sync) * Otherwise check if either cpus are near enough in load to allow this * task to be woken on this_cpu. */ - balanced = !this_load || - 100*(this_load + effective_load(tg, this_cpu, weight, weight)) <= - imbalance*(load + effective_load(tg, prev_cpu, 0, weight)); + if (this_load) { + unsigned long this_eff_load, prev_eff_load; + + this_eff_load = 100; + this_eff_load *= power_of(prev_cpu); + this_eff_load *= this_load + + effective_load(tg, this_cpu, weight, weight); + + prev_eff_load = 100 + (sd->imbalance_pct - 100) / 2; + prev_eff_load *= power_of(this_cpu); + prev_eff_load *= load + effective_load(tg, prev_cpu, 0, weight); + + balanced = this_eff_load <= prev_eff_load; + } else + balanced = true; + rcu_read_unlock(); /* * If the currently running task will sleep within @@ -1306,7 +1286,7 @@ static int wake_affine(struct sched_domain *sd, struct task_struct *p, int sync) if (sync && balanced) return 1; - schedstat_inc(p, se.nr_wakeups_affine_attempts); + schedstat_inc(p, se.statistics.nr_wakeups_affine_attempts); tl_per_task = cpu_avg_load_per_task(this_cpu); if (balanced || @@ -1318,7 +1298,7 @@ static int wake_affine(struct sched_domain *sd, struct task_struct *p, int sync) * there is no bad imbalance. */ schedstat_inc(sd, ttwu_move_affine); - schedstat_inc(p, se.nr_wakeups_affine); + schedstat_inc(p, se.statistics.nr_wakeups_affine); return 1; } @@ -1406,29 +1386,48 @@ find_idlest_cpu(struct sched_group *group, struct task_struct *p, int this_cpu) /* * Try and locate an idle CPU in the sched_domain. */ -static int -select_idle_sibling(struct task_struct *p, struct sched_domain *sd, int target) +static int select_idle_sibling(struct task_struct *p, int target) { int cpu = smp_processor_id(); int prev_cpu = task_cpu(p); + struct sched_domain *sd; int i; /* - * If this domain spans both cpu and prev_cpu (see the SD_WAKE_AFFINE - * test in select_task_rq_fair) and the prev_cpu is idle then that's - * always a better target than the current cpu. + * If the task is going to be woken-up on this cpu and if it is + * already idle, then it is the right target. + */ + if (target == cpu && idle_cpu(cpu)) + return cpu; + + /* + * If the task is going to be woken-up on the cpu where it previously + * ran and if it is currently idle, then it the right target. */ - if (target == cpu && !cpu_rq(prev_cpu)->cfs.nr_running) + if (target == prev_cpu && idle_cpu(prev_cpu)) return prev_cpu; /* - * Otherwise, iterate the domain and find an elegible idle cpu. + * Otherwise, iterate the domains and find an elegible idle cpu. */ - for_each_cpu_and(i, sched_domain_span(sd), &p->cpus_allowed) { - if (!cpu_rq(i)->cfs.nr_running) { - target = i; + for_each_domain(target, sd) { + if (!(sd->flags & SD_SHARE_PKG_RESOURCES)) break; + + for_each_cpu_and(i, sched_domain_span(sd), &p->cpus_allowed) { + if (idle_cpu(i)) { + target = i; + break; + } } + + /* + * Lets stop looking for an idle sibling when we reached + * the domain that spans the current cpu and prev_cpu. + */ + if (cpumask_test_cpu(cpu, sched_domain_span(sd)) && + cpumask_test_cpu(prev_cpu, sched_domain_span(sd))) + break; } return target; @@ -1445,7 +1444,8 @@ select_idle_sibling(struct task_struct *p, struct sched_domain *sd, int target) * * preempt must be disabled. */ -static int select_task_rq_fair(struct task_struct *p, int sd_flag, int wake_flags) +static int +select_task_rq_fair(struct rq *rq, struct task_struct *p, int sd_flag, int wake_flags) { struct sched_domain *tmp, *affine_sd = NULL, *sd = NULL; int cpu = smp_processor_id(); @@ -1456,8 +1456,7 @@ static int select_task_rq_fair(struct task_struct *p, int sd_flag, int wake_flag int sync = wake_flags & WF_SYNC; if (sd_flag & SD_BALANCE_WAKE) { - if (sched_feat(AFFINE_WAKEUPS) && - cpumask_test_cpu(cpu, &p->cpus_allowed)) + if (cpumask_test_cpu(cpu, &p->cpus_allowed)) want_affine = 1; new_cpu = prev_cpu; } @@ -1491,34 +1490,13 @@ static int select_task_rq_fair(struct task_struct *p, int sd_flag, int wake_flag } /* - * While iterating the domains looking for a spanning - * WAKE_AFFINE domain, adjust the affine target to any idle cpu - * in cache sharing domains along the way. + * If both cpu and prev_cpu are part of this domain, + * cpu is a valid SD_WAKE_AFFINE target. */ - if (want_affine) { - int target = -1; - - /* - * If both cpu and prev_cpu are part of this domain, - * cpu is a valid SD_WAKE_AFFINE target. - */ - if (cpumask_test_cpu(prev_cpu, sched_domain_span(tmp))) - target = cpu; - - /* - * If there's an idle sibling in this domain, make that - * the wake_affine target instead of the current cpu. - */ - if (tmp->flags & SD_SHARE_PKG_RESOURCES) - target = select_idle_sibling(p, tmp, target); - - if (target >= 0) { - if (tmp->flags & SD_WAKE_AFFINE) { - affine_sd = tmp; - want_affine = 0; - } - cpu = target; - } + if (want_affine && (tmp->flags & SD_WAKE_AFFINE) && + cpumask_test_cpu(prev_cpu, sched_domain_span(tmp))) { + affine_sd = tmp; + want_affine = 0; } if (!want_sd && !want_affine) @@ -1531,22 +1509,29 @@ static int select_task_rq_fair(struct task_struct *p, int sd_flag, int wake_flag sd = tmp; } +#ifdef CONFIG_FAIR_GROUP_SCHED if (sched_feat(LB_SHARES_UPDATE)) { /* * Pick the largest domain to update shares over */ tmp = sd; - if (affine_sd && (!tmp || - cpumask_weight(sched_domain_span(affine_sd)) > - cpumask_weight(sched_domain_span(sd)))) + if (affine_sd && (!tmp || affine_sd->span_weight > sd->span_weight)) tmp = affine_sd; - if (tmp) + if (tmp) { + raw_spin_unlock(&rq->lock); update_shares(tmp); + raw_spin_lock(&rq->lock); + } } +#endif - if (affine_sd && wake_affine(affine_sd, p, sync)) - return cpu; + if (affine_sd) { + if (cpu == prev_cpu || wake_affine(affine_sd, p, sync)) + return select_idle_sibling(p, cpu); + else + return select_idle_sibling(p, prev_cpu); + } while (sd) { int load_idx = sd->forkexec_idx; @@ -1576,10 +1561,10 @@ static int select_task_rq_fair(struct task_struct *p, int sd_flag, int wake_flag /* Now try balancing at a lower domain level of new_cpu */ cpu = new_cpu; - weight = cpumask_weight(sched_domain_span(sd)); + weight = sd->span_weight; sd = NULL; for_each_domain(cpu, tmp) { - if (weight <= cpumask_weight(sched_domain_span(tmp))) + if (weight <= tmp->span_weight) break; if (tmp->flags & sd_flag) sd = tmp; @@ -1591,63 +1576,26 @@ static int select_task_rq_fair(struct task_struct *p, int sd_flag, int wake_flag } #endif /* CONFIG_SMP */ -/* - * Adaptive granularity - * - * se->avg_wakeup gives the average time a task runs until it does a wakeup, - * with the limit of wakeup_gran -- when it never does a wakeup. - * - * So the smaller avg_wakeup is the faster we want this task to preempt, - * but we don't want to treat the preemptee unfairly and therefore allow it - * to run for at least the amount of time we'd like to run. - * - * NOTE: we use 2*avg_wakeup to increase the probability of actually doing one - * - * NOTE: we use *nr_running to scale with load, this nicely matches the - * degrading latency on load. - */ -static unsigned long -adaptive_gran(struct sched_entity *curr, struct sched_entity *se) -{ - u64 this_run = curr->sum_exec_runtime - curr->prev_sum_exec_runtime; - u64 expected_wakeup = 2*se->avg_wakeup * cfs_rq_of(se)->nr_running; - u64 gran = 0; - - if (this_run < expected_wakeup) - gran = expected_wakeup - this_run; - - return min_t(s64, gran, sysctl_sched_wakeup_granularity); -} - static unsigned long wakeup_gran(struct sched_entity *curr, struct sched_entity *se) { unsigned long gran = sysctl_sched_wakeup_granularity; - if (cfs_rq_of(curr)->curr && sched_feat(ADAPTIVE_GRAN)) - gran = adaptive_gran(curr, se); - /* * Since its curr running now, convert the gran from real-time * to virtual-time in his units. + * + * By using 'se' instead of 'curr' we penalize light tasks, so + * they get preempted easier. That is, if 'se' < 'curr' then + * the resulting gran will be larger, therefore penalizing the + * lighter, if otoh 'se' > 'curr' then the resulting gran will + * be smaller, again penalizing the lighter task. + * + * This is especially important for buddies when the leftmost + * task is higher priority than the buddy. */ - if (sched_feat(ASYM_GRAN)) { - /* - * By using 'se' instead of 'curr' we penalize light tasks, so - * they get preempted easier. That is, if 'se' < 'curr' then - * the resulting gran will be larger, therefore penalizing the - * lighter, if otoh 'se' > 'curr' then the resulting gran will - * be smaller, again penalizing the lighter task. - * - * This is especially important for buddies when the leftmost - * task is higher priority than the buddy. - */ - if (unlikely(se->load.weight != NICE_0_LOAD)) - gran = calc_delta_fair(gran, se); - } else { - if (unlikely(curr->load.weight != NICE_0_LOAD)) - gran = calc_delta_fair(gran, curr); - } + if (unlikely(se->load.weight != NICE_0_LOAD)) + gran = calc_delta_fair(gran, se); return gran; } @@ -1705,7 +1653,6 @@ static void check_preempt_wakeup(struct rq *rq, struct task_struct *p, int wake_ struct task_struct *curr = rq->curr; struct sched_entity *se = &curr->se, *pse = &p->se; struct cfs_rq *cfs_rq = task_cfs_rq(curr); - int sync = wake_flags & WF_SYNC; int scale = cfs_rq->nr_running >= sched_nr_latency; if (unlikely(rt_prio(p->prio))) @@ -1738,14 +1685,6 @@ static void check_preempt_wakeup(struct rq *rq, struct task_struct *p, int wake_ if (unlikely(curr->policy == SCHED_IDLE)) goto preempt; - if (sched_feat(WAKEUP_SYNC) && sync) - goto preempt; - - if (sched_feat(WAKEUP_OVERLAP) && - se->avg_overlap < sysctl_sched_migration_cost && - pse->avg_overlap < sysctl_sched_migration_cost) - goto preempt; - if (!sched_feat(WAKEUP_PREEMPT)) return; @@ -1844,13 +1783,13 @@ int can_migrate_task(struct task_struct *p, struct rq *rq, int this_cpu, * 3) are cache-hot on their current CPU. */ if (!cpumask_test_cpu(this_cpu, &p->cpus_allowed)) { - schedstat_inc(p, se.nr_failed_migrations_affine); + schedstat_inc(p, se.statistics.nr_failed_migrations_affine); return 0; } *all_pinned = 0; if (task_running(rq, p)) { - schedstat_inc(p, se.nr_failed_migrations_running); + schedstat_inc(p, se.statistics.nr_failed_migrations_running); return 0; } @@ -1866,14 +1805,14 @@ int can_migrate_task(struct task_struct *p, struct rq *rq, int this_cpu, #ifdef CONFIG_SCHEDSTATS if (tsk_cache_hot) { schedstat_inc(sd, lb_hot_gained[idle]); - schedstat_inc(p, se.nr_forced_migrations); + schedstat_inc(p, se.statistics.nr_forced_migrations); } #endif return 1; } if (tsk_cache_hot) { - schedstat_inc(p, se.nr_failed_migrations_hot); + schedstat_inc(p, se.statistics.nr_failed_migrations_hot); return 0; } return 1; @@ -2311,7 +2250,7 @@ unsigned long __weak arch_scale_freq_power(struct sched_domain *sd, int cpu) unsigned long default_scale_smt_power(struct sched_domain *sd, int cpu) { - unsigned long weight = cpumask_weight(sched_domain_span(sd)); + unsigned long weight = sd->span_weight; unsigned long smt_gain = sd->smt_gain; smt_gain /= weight; @@ -2344,17 +2283,10 @@ unsigned long scale_rt_power(int cpu) static void update_cpu_power(struct sched_domain *sd, int cpu) { - unsigned long weight = cpumask_weight(sched_domain_span(sd)); + unsigned long weight = sd->span_weight; unsigned long power = SCHED_LOAD_SCALE; struct sched_group *sdg = sd->groups; - if (sched_feat(ARCH_POWER)) - power *= arch_scale_freq_power(sd, cpu); - else - power *= default_scale_freq_power(sd, cpu); - - power >>= SCHED_LOAD_SHIFT; - if ((sd->flags & SD_SHARE_CPUPOWER) && weight > 1) { if (sched_feat(ARCH_POWER)) power *= arch_scale_smt_power(sd, cpu); @@ -2364,12 +2296,22 @@ static void update_cpu_power(struct sched_domain *sd, int cpu) power >>= SCHED_LOAD_SHIFT; } + sdg->cpu_power_orig = power; + + if (sched_feat(ARCH_POWER)) + power *= arch_scale_freq_power(sd, cpu); + else + power *= default_scale_freq_power(sd, cpu); + + power >>= SCHED_LOAD_SHIFT; + power *= scale_rt_power(cpu); power >>= SCHED_LOAD_SHIFT; if (!power) power = 1; + cpu_rq(cpu)->cpu_power = power; sdg->cpu_power = power; } @@ -2395,6 +2337,31 @@ static void update_group_power(struct sched_domain *sd, int cpu) sdg->cpu_power = power; } +/* + * Try and fix up capacity for tiny siblings, this is needed when + * things like SD_ASYM_PACKING need f_b_g to select another sibling + * which on its own isn't powerful enough. + * + * See update_sd_pick_busiest() and check_asym_packing(). + */ +static inline int +fix_small_capacity(struct sched_domain *sd, struct sched_group *group) +{ + /* + * Only siblings can have significantly less than SCHED_LOAD_SCALE + */ + if (sd->level != SD_LV_SIBLING) + return 0; + + /* + * If ~90% of the cpu_power is still there, we're good. + */ + if (group->cpu_power * 32 > group->cpu_power_orig * 29) + return 1; + + return 0; +} + /** * update_sg_lb_stats - Update sched_group's statistics for load balancing. * @sd: The sched_domain whose statistics are to be updated. @@ -2460,14 +2427,14 @@ static inline void update_sg_lb_stats(struct sched_domain *sd, * domains. In the newly idle case, we will allow all the cpu's * to do the newly idle load balance. */ - if (idle != CPU_NEWLY_IDLE && local_group && - balance_cpu != this_cpu) { - *balance = 0; - return; + if (idle != CPU_NEWLY_IDLE && local_group) { + if (balance_cpu != this_cpu) { + *balance = 0; + return; + } + update_group_power(sd, this_cpu); } - update_group_power(sd, this_cpu); - /* Adjust by relative CPU power of the group */ sgs->avg_load = (sgs->group_load * SCHED_LOAD_SCALE) / group->cpu_power; @@ -2488,6 +2455,51 @@ static inline void update_sg_lb_stats(struct sched_domain *sd, sgs->group_capacity = DIV_ROUND_CLOSEST(group->cpu_power, SCHED_LOAD_SCALE); + if (!sgs->group_capacity) + sgs->group_capacity = fix_small_capacity(sd, group); +} + +/** + * update_sd_pick_busiest - return 1 on busiest group + * @sd: sched_domain whose statistics are to be checked + * @sds: sched_domain statistics + * @sg: sched_group candidate to be checked for being the busiest + * @sgs: sched_group statistics + * @this_cpu: the current cpu + * + * Determine if @sg is a busier group than the previously selected + * busiest group. + */ +static bool update_sd_pick_busiest(struct sched_domain *sd, + struct sd_lb_stats *sds, + struct sched_group *sg, + struct sg_lb_stats *sgs, + int this_cpu) +{ + if (sgs->avg_load <= sds->max_load) + return false; + + if (sgs->sum_nr_running > sgs->group_capacity) + return true; + + if (sgs->group_imb) + return true; + + /* + * ASYM_PACKING needs to move all the work to the lowest + * numbered CPUs in the group, therefore mark all groups + * higher than ourself as busy. + */ + if ((sd->flags & SD_ASYM_PACKING) && sgs->sum_nr_running && + this_cpu < group_first_cpu(sg)) { + if (!sds->busiest) + return true; + + if (group_first_cpu(sds->busiest) > group_first_cpu(sg)) + return true; + } + + return false; } /** @@ -2495,7 +2507,7 @@ static inline void update_sg_lb_stats(struct sched_domain *sd, * @sd: sched_domain whose statistics are to be updated. * @this_cpu: Cpu for which load balance is currently performed. * @idle: Idle status of this_cpu - * @sd_idle: Idle status of the sched_domain containing group. + * @sd_idle: Idle status of the sched_domain containing sg. * @cpus: Set of cpus considered for load balancing. * @balance: Should we balance. * @sds: variable to hold the statistics for this sched_domain. @@ -2506,7 +2518,7 @@ static inline void update_sd_lb_stats(struct sched_domain *sd, int this_cpu, struct sd_lb_stats *sds) { struct sched_domain *child = sd->child; - struct sched_group *group = sd->groups; + struct sched_group *sg = sd->groups; struct sg_lb_stats sgs; int load_idx, prefer_sibling = 0; @@ -2519,21 +2531,20 @@ static inline void update_sd_lb_stats(struct sched_domain *sd, int this_cpu, do { int local_group; - local_group = cpumask_test_cpu(this_cpu, - sched_group_cpus(group)); + local_group = cpumask_test_cpu(this_cpu, sched_group_cpus(sg)); memset(&sgs, 0, sizeof(sgs)); - update_sg_lb_stats(sd, group, this_cpu, idle, load_idx, sd_idle, + update_sg_lb_stats(sd, sg, this_cpu, idle, load_idx, sd_idle, local_group, cpus, balance, &sgs); if (local_group && !(*balance)) return; sds->total_load += sgs.group_load; - sds->total_pwr += group->cpu_power; + sds->total_pwr += sg->cpu_power; /* * In case the child domain prefers tasks go to siblings - * first, lower the group capacity to one so that we'll try + * first, lower the sg capacity to one so that we'll try * and move all the excess tasks away. */ if (prefer_sibling) @@ -2541,23 +2552,72 @@ static inline void update_sd_lb_stats(struct sched_domain *sd, int this_cpu, if (local_group) { sds->this_load = sgs.avg_load; - sds->this = group; + sds->this = sg; sds->this_nr_running = sgs.sum_nr_running; sds->this_load_per_task = sgs.sum_weighted_load; - } else if (sgs.avg_load > sds->max_load && - (sgs.sum_nr_running > sgs.group_capacity || - sgs.group_imb)) { + } else if (update_sd_pick_busiest(sd, sds, sg, &sgs, this_cpu)) { sds->max_load = sgs.avg_load; - sds->busiest = group; + sds->busiest = sg; sds->busiest_nr_running = sgs.sum_nr_running; sds->busiest_group_capacity = sgs.group_capacity; sds->busiest_load_per_task = sgs.sum_weighted_load; sds->group_imb = sgs.group_imb; } - update_sd_power_savings_stats(group, sds, local_group, &sgs); - group = group->next; - } while (group != sd->groups); + update_sd_power_savings_stats(sg, sds, local_group, &sgs); + sg = sg->next; + } while (sg != sd->groups); +} + +int __weak arch_sd_sibling_asym_packing(void) +{ + return 0*SD_ASYM_PACKING; +} + +/** + * check_asym_packing - Check to see if the group is packed into the + * sched doman. + * + * This is primarily intended to used at the sibling level. Some + * cores like POWER7 prefer to use lower numbered SMT threads. In the + * case of POWER7, it can move to lower SMT modes only when higher + * threads are idle. When in lower SMT modes, the threads will + * perform better since they share less core resources. Hence when we + * have idle threads, we want them to be the higher ones. + * + * This packing function is run on idle threads. It checks to see if + * the busiest CPU in this domain (core in the P7 case) has a higher + * CPU number than the packing function is being run on. Here we are + * assuming lower CPU number will be equivalent to lower a SMT thread + * number. + * + * Returns 1 when packing is required and a task should be moved to + * this CPU. The amount of the imbalance is returned in *imbalance. + * + * @sd: The sched_domain whose packing is to be checked. + * @sds: Statistics of the sched_domain which is to be packed + * @this_cpu: The cpu at whose sched_domain we're performing load-balance. + * @imbalance: returns amount of imbalanced due to packing. + */ +static int check_asym_packing(struct sched_domain *sd, + struct sd_lb_stats *sds, + int this_cpu, unsigned long *imbalance) +{ + int busiest_cpu; + + if (!(sd->flags & SD_ASYM_PACKING)) + return 0; + + if (!sds->busiest) + return 0; + + busiest_cpu = group_first_cpu(sds->busiest); + if (this_cpu > busiest_cpu) + return 0; + + *imbalance = DIV_ROUND_CLOSEST(sds->max_load * sds->busiest->cpu_power, + SCHED_LOAD_SCALE); + return 1; } /** @@ -2752,6 +2812,10 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, if (!(*balance)) goto ret; + if ((idle == CPU_IDLE || idle == CPU_NEWLY_IDLE) && + check_asym_packing(sd, &sds, this_cpu, imbalance)) + return sds.busiest; + if (!sds.busiest || sds.busiest_nr_running == 0) goto out_balanced; @@ -2786,8 +2850,9 @@ ret: * find_busiest_queue - find the busiest runqueue among the cpus in group. */ static struct rq * -find_busiest_queue(struct sched_group *group, enum cpu_idle_type idle, - unsigned long imbalance, const struct cpumask *cpus) +find_busiest_queue(struct sched_domain *sd, struct sched_group *group, + enum cpu_idle_type idle, unsigned long imbalance, + const struct cpumask *cpus) { struct rq *busiest = NULL, *rq; unsigned long max_load = 0; @@ -2798,6 +2863,9 @@ find_busiest_queue(struct sched_group *group, enum cpu_idle_type idle, unsigned long capacity = DIV_ROUND_CLOSEST(power, SCHED_LOAD_SCALE); unsigned long wl; + if (!capacity) + capacity = fix_small_capacity(sd, group); + if (!cpumask_test_cpu(i, cpus)) continue; @@ -2837,9 +2905,19 @@ find_busiest_queue(struct sched_group *group, enum cpu_idle_type idle, /* Working cpumask for load_balance and load_balance_newidle. */ static DEFINE_PER_CPU(cpumask_var_t, load_balance_tmpmask); -static int need_active_balance(struct sched_domain *sd, int sd_idle, int idle) +static int need_active_balance(struct sched_domain *sd, int sd_idle, int idle, + int busiest_cpu, int this_cpu) { if (idle == CPU_NEWLY_IDLE) { + + /* + * ASYM_PACKING needs to force migrate tasks from busy but + * higher numbered CPUs in order to pack all tasks in the + * lowest numbered CPUs. + */ + if ((sd->flags & SD_ASYM_PACKING) && busiest_cpu > this_cpu) + return 1; + /* * The only task running in a non-idle cpu can be moved to this * cpu in an attempt to completely freeup the other CPU @@ -2870,6 +2948,8 @@ static int need_active_balance(struct sched_domain *sd, int sd_idle, int idle) return unlikely(sd->nr_balance_failed > sd->cache_nice_tries+2); } +static int active_load_balance_cpu_stop(void *data); + /* * Check this_cpu to ensure it is balanced within domain. Attempt to move * tasks if there is an imbalance. @@ -2912,7 +2992,7 @@ redo: goto out_balanced; } - busiest = find_busiest_queue(group, idle, imbalance, cpus); + busiest = find_busiest_queue(sd, group, idle, imbalance, cpus); if (!busiest) { schedstat_inc(sd, lb_nobusyq[idle]); goto out_balanced; @@ -2956,11 +3036,13 @@ redo: schedstat_inc(sd, lb_failed[idle]); sd->nr_balance_failed++; - if (need_active_balance(sd, sd_idle, idle)) { + if (need_active_balance(sd, sd_idle, idle, cpu_of(busiest), + this_cpu)) { raw_spin_lock_irqsave(&busiest->lock, flags); - /* don't kick the migration_thread, if the curr - * task on busiest cpu can't be moved to this_cpu + /* don't kick the active_load_balance_cpu_stop, + * if the curr task on busiest cpu can't be + * moved to this_cpu */ if (!cpumask_test_cpu(this_cpu, &busiest->curr->cpus_allowed)) { @@ -2970,14 +3052,22 @@ redo: goto out_one_pinned; } + /* + * ->active_balance synchronizes accesses to + * ->active_balance_work. Once set, it's cleared + * only after active load balance is finished. + */ if (!busiest->active_balance) { busiest->active_balance = 1; busiest->push_cpu = this_cpu; active_balance = 1; } raw_spin_unlock_irqrestore(&busiest->lock, flags); + if (active_balance) - wake_up_process(busiest->migration_thread); + stop_one_cpu_nowait(cpu_of(busiest), + active_load_balance_cpu_stop, busiest, + &busiest->active_balance_work); /* * We've kicked active balancing, reset the failure @@ -3084,24 +3174,29 @@ static void idle_balance(int this_cpu, struct rq *this_rq) } /* - * active_load_balance is run by migration threads. It pushes running tasks - * off the busiest CPU onto idle CPUs. It requires at least 1 task to be - * running on each physical CPU where possible, and avoids physical / - * logical imbalances. - * - * Called with busiest_rq locked. + * active_load_balance_cpu_stop is run by cpu stopper. It pushes + * running tasks off the busiest CPU onto idle CPUs. It requires at + * least 1 task to be running on each physical CPU where possible, and + * avoids physical / logical imbalances. */ -static void active_load_balance(struct rq *busiest_rq, int busiest_cpu) +static int active_load_balance_cpu_stop(void *data) { + struct rq *busiest_rq = data; + int busiest_cpu = cpu_of(busiest_rq); int target_cpu = busiest_rq->push_cpu; + struct rq *target_rq = cpu_rq(target_cpu); struct sched_domain *sd; - struct rq *target_rq; + + raw_spin_lock_irq(&busiest_rq->lock); + + /* make sure the requested cpu hasn't gone down in the meantime */ + if (unlikely(busiest_cpu != smp_processor_id() || + !busiest_rq->active_balance)) + goto out_unlock; /* Is there any task to move? */ if (busiest_rq->nr_running <= 1) - return; - - target_rq = cpu_rq(target_cpu); + goto out_unlock; /* * This condition is "impossible", if it occurs @@ -3112,8 +3207,6 @@ static void active_load_balance(struct rq *busiest_rq, int busiest_cpu) /* move a task from busiest_rq to target_rq */ double_lock_balance(busiest_rq, target_rq); - update_rq_clock(busiest_rq); - update_rq_clock(target_rq); /* Search for an sd spanning us and the target CPU. */ for_each_domain(target_cpu, sd) { @@ -3132,16 +3225,47 @@ static void active_load_balance(struct rq *busiest_rq, int busiest_cpu) schedstat_inc(sd, alb_failed); } double_unlock_balance(busiest_rq, target_rq); +out_unlock: + busiest_rq->active_balance = 0; + raw_spin_unlock_irq(&busiest_rq->lock); + return 0; } #ifdef CONFIG_NO_HZ + +static DEFINE_PER_CPU(struct call_single_data, remote_sched_softirq_cb); + +static void trigger_sched_softirq(void *data) +{ + raise_softirq_irqoff(SCHED_SOFTIRQ); +} + +static inline void init_sched_softirq_csd(struct call_single_data *csd) +{ + csd->func = trigger_sched_softirq; + csd->info = NULL; + csd->flags = 0; + csd->priv = 0; +} + +/* + * idle load balancing details + * - One of the idle CPUs nominates itself as idle load_balancer, while + * entering idle. + * - This idle load balancer CPU will also go into tickless mode when + * it is idle, just like all other idle CPUs + * - When one of the busy CPUs notice that there may be an idle rebalancing + * needed, they will kick the idle load balancer, which then does idle + * load balancing for all the idle CPUs. + */ static struct { atomic_t load_balancer; - cpumask_var_t cpu_mask; - cpumask_var_t ilb_grp_nohz_mask; -} nohz ____cacheline_aligned = { - .load_balancer = ATOMIC_INIT(-1), -}; + atomic_t first_pick_cpu; + atomic_t second_pick_cpu; + cpumask_var_t idle_cpus_mask; + cpumask_var_t grp_idle_mask; + unsigned long next_balance; /* in jiffy units */ +} nohz ____cacheline_aligned; int get_nohz_load_balancer(void) { @@ -3195,17 +3319,17 @@ static inline struct sched_domain *lowest_flag_domain(int cpu, int flag) */ static inline int is_semi_idle_group(struct sched_group *ilb_group) { - cpumask_and(nohz.ilb_grp_nohz_mask, nohz.cpu_mask, + cpumask_and(nohz.grp_idle_mask, nohz.idle_cpus_mask, sched_group_cpus(ilb_group)); /* * A sched_group is semi-idle when it has atleast one busy cpu * and atleast one idle cpu. */ - if (cpumask_empty(nohz.ilb_grp_nohz_mask)) + if (cpumask_empty(nohz.grp_idle_mask)) return 0; - if (cpumask_equal(nohz.ilb_grp_nohz_mask, sched_group_cpus(ilb_group))) + if (cpumask_equal(nohz.grp_idle_mask, sched_group_cpus(ilb_group))) return 0; return 1; @@ -3238,7 +3362,7 @@ static int find_new_ilb(int cpu) * Optimize for the case when we have no idle CPUs or only one * idle CPU. Don't walk the sched_domain hierarchy in such cases */ - if (cpumask_weight(nohz.cpu_mask) < 2) + if (cpumask_weight(nohz.idle_cpus_mask) < 2) goto out_done; for_each_flag_domain(cpu, sd, SD_POWERSAVINGS_BALANCE) { @@ -3246,7 +3370,7 @@ static int find_new_ilb(int cpu) do { if (is_semi_idle_group(ilb_group)) - return cpumask_first(nohz.ilb_grp_nohz_mask); + return cpumask_first(nohz.grp_idle_mask); ilb_group = ilb_group->next; @@ -3254,98 +3378,116 @@ static int find_new_ilb(int cpu) } out_done: - return cpumask_first(nohz.cpu_mask); + return nr_cpu_ids; } #else /* (CONFIG_SCHED_MC || CONFIG_SCHED_SMT) */ static inline int find_new_ilb(int call_cpu) { - return cpumask_first(nohz.cpu_mask); + return nr_cpu_ids; } #endif /* + * Kick a CPU to do the nohz balancing, if it is time for it. We pick the + * nohz_load_balancer CPU (if there is one) otherwise fallback to any idle + * CPU (if there is one). + */ +static void nohz_balancer_kick(int cpu) +{ + int ilb_cpu; + + nohz.next_balance++; + + ilb_cpu = get_nohz_load_balancer(); + + if (ilb_cpu >= nr_cpu_ids) { + ilb_cpu = cpumask_first(nohz.idle_cpus_mask); + if (ilb_cpu >= nr_cpu_ids) + return; + } + + if (!cpu_rq(ilb_cpu)->nohz_balance_kick) { + struct call_single_data *cp; + + cpu_rq(ilb_cpu)->nohz_balance_kick = 1; + cp = &per_cpu(remote_sched_softirq_cb, cpu); + __smp_call_function_single(ilb_cpu, cp, 0); + } + return; +} + +/* * This routine will try to nominate the ilb (idle load balancing) * owner among the cpus whose ticks are stopped. ilb owner will do the idle - * load balancing on behalf of all those cpus. If all the cpus in the system - * go into this tickless mode, then there will be no ilb owner (as there is - * no need for one) and all the cpus will sleep till the next wakeup event - * arrives... - * - * For the ilb owner, tick is not stopped. And this tick will be used - * for idle load balancing. ilb owner will still be part of - * nohz.cpu_mask.. + * load balancing on behalf of all those cpus. * - * While stopping the tick, this cpu will become the ilb owner if there - * is no other owner. And will be the owner till that cpu becomes busy - * or if all cpus in the system stop their ticks at which point - * there is no need for ilb owner. + * When the ilb owner becomes busy, we will not have new ilb owner until some + * idle CPU wakes up and goes back to idle or some busy CPU tries to kick + * idle load balancing by kicking one of the idle CPUs. * - * When the ilb owner becomes busy, it nominates another owner, during the - * next busy scheduler_tick() + * Ticks are stopped for the ilb owner as well, with busy CPU kicking this + * ilb owner CPU in future (when there is a need for idle load balancing on + * behalf of all idle CPUs). */ -int select_nohz_load_balancer(int stop_tick) +void select_nohz_load_balancer(int stop_tick) { int cpu = smp_processor_id(); if (stop_tick) { - cpu_rq(cpu)->in_nohz_recently = 1; - if (!cpu_active(cpu)) { if (atomic_read(&nohz.load_balancer) != cpu) - return 0; + return; /* * If we are going offline and still the leader, * give up! */ - if (atomic_cmpxchg(&nohz.load_balancer, cpu, -1) != cpu) + if (atomic_cmpxchg(&nohz.load_balancer, cpu, + nr_cpu_ids) != cpu) BUG(); - return 0; + return; } - cpumask_set_cpu(cpu, nohz.cpu_mask); + cpumask_set_cpu(cpu, nohz.idle_cpus_mask); - /* time for ilb owner also to sleep */ - if (cpumask_weight(nohz.cpu_mask) == num_active_cpus()) { - if (atomic_read(&nohz.load_balancer) == cpu) - atomic_set(&nohz.load_balancer, -1); - return 0; - } + if (atomic_read(&nohz.first_pick_cpu) == cpu) + atomic_cmpxchg(&nohz.first_pick_cpu, cpu, nr_cpu_ids); + if (atomic_read(&nohz.second_pick_cpu) == cpu) + atomic_cmpxchg(&nohz.second_pick_cpu, cpu, nr_cpu_ids); - if (atomic_read(&nohz.load_balancer) == -1) { - /* make me the ilb owner */ - if (atomic_cmpxchg(&nohz.load_balancer, -1, cpu) == -1) - return 1; - } else if (atomic_read(&nohz.load_balancer) == cpu) { + if (atomic_read(&nohz.load_balancer) >= nr_cpu_ids) { int new_ilb; - if (!(sched_smt_power_savings || - sched_mc_power_savings)) - return 1; + /* make me the ilb owner */ + if (atomic_cmpxchg(&nohz.load_balancer, nr_cpu_ids, + cpu) != nr_cpu_ids) + return; + /* * Check to see if there is a more power-efficient * ilb. */ new_ilb = find_new_ilb(cpu); if (new_ilb < nr_cpu_ids && new_ilb != cpu) { - atomic_set(&nohz.load_balancer, -1); + atomic_set(&nohz.load_balancer, nr_cpu_ids); resched_cpu(new_ilb); - return 0; + return; } - return 1; + return; } } else { - if (!cpumask_test_cpu(cpu, nohz.cpu_mask)) - return 0; + if (!cpumask_test_cpu(cpu, nohz.idle_cpus_mask)) + return; - cpumask_clear_cpu(cpu, nohz.cpu_mask); + cpumask_clear_cpu(cpu, nohz.idle_cpus_mask); if (atomic_read(&nohz.load_balancer) == cpu) - if (atomic_cmpxchg(&nohz.load_balancer, cpu, -1) != cpu) + if (atomic_cmpxchg(&nohz.load_balancer, cpu, + nr_cpu_ids) != cpu) BUG(); } - return 0; + return; } #endif @@ -3427,11 +3569,102 @@ out: rq->next_balance = next_balance; } +#ifdef CONFIG_NO_HZ /* - * run_rebalance_domains is triggered when needed from the scheduler tick. - * In CONFIG_NO_HZ case, the idle load balance owner will do the + * In CONFIG_NO_HZ case, the idle balance kickee will do the * rebalancing for all the cpus for whom scheduler ticks are stopped. */ +static void nohz_idle_balance(int this_cpu, enum cpu_idle_type idle) +{ + struct rq *this_rq = cpu_rq(this_cpu); + struct rq *rq; + int balance_cpu; + + if (idle != CPU_IDLE || !this_rq->nohz_balance_kick) + return; + + for_each_cpu(balance_cpu, nohz.idle_cpus_mask) { + if (balance_cpu == this_cpu) + continue; + + /* + * If this cpu gets work to do, stop the load balancing + * work being done for other cpus. Next load + * balancing owner will pick it up. + */ + if (need_resched()) { + this_rq->nohz_balance_kick = 0; + break; + } + + raw_spin_lock_irq(&this_rq->lock); + update_rq_clock(this_rq); + update_cpu_load(this_rq); + raw_spin_unlock_irq(&this_rq->lock); + + rebalance_domains(balance_cpu, CPU_IDLE); + + rq = cpu_rq(balance_cpu); + if (time_after(this_rq->next_balance, rq->next_balance)) + this_rq->next_balance = rq->next_balance; + } + nohz.next_balance = this_rq->next_balance; + this_rq->nohz_balance_kick = 0; +} + +/* + * Current heuristic for kicking the idle load balancer + * - first_pick_cpu is the one of the busy CPUs. It will kick + * idle load balancer when it has more than one process active. This + * eliminates the need for idle load balancing altogether when we have + * only one running process in the system (common case). + * - If there are more than one busy CPU, idle load balancer may have + * to run for active_load_balance to happen (i.e., two busy CPUs are + * SMT or core siblings and can run better if they move to different + * physical CPUs). So, second_pick_cpu is the second of the busy CPUs + * which will kick idle load balancer as soon as it has any load. + */ +static inline int nohz_kick_needed(struct rq *rq, int cpu) +{ + unsigned long now = jiffies; + int ret; + int first_pick_cpu, second_pick_cpu; + + if (time_before(now, nohz.next_balance)) + return 0; + + if (!rq->nr_running) + return 0; + + first_pick_cpu = atomic_read(&nohz.first_pick_cpu); + second_pick_cpu = atomic_read(&nohz.second_pick_cpu); + + if (first_pick_cpu < nr_cpu_ids && first_pick_cpu != cpu && + second_pick_cpu < nr_cpu_ids && second_pick_cpu != cpu) + return 0; + + ret = atomic_cmpxchg(&nohz.first_pick_cpu, nr_cpu_ids, cpu); + if (ret == nr_cpu_ids || ret == cpu) { + atomic_cmpxchg(&nohz.second_pick_cpu, cpu, nr_cpu_ids); + if (rq->nr_running > 1) + return 1; + } else { + ret = atomic_cmpxchg(&nohz.second_pick_cpu, nr_cpu_ids, cpu); + if (ret == nr_cpu_ids || ret == cpu) { + if (rq->nr_running) + return 1; + } + } + return 0; +} +#else +static void nohz_idle_balance(int this_cpu, enum cpu_idle_type idle) { } +#endif + +/* + * run_rebalance_domains is triggered when needed from the scheduler tick. + * Also triggered for nohz idle balancing (with nohz_balancing_kick set). + */ static void run_rebalance_domains(struct softirq_action *h) { int this_cpu = smp_processor_id(); @@ -3441,37 +3674,12 @@ static void run_rebalance_domains(struct softirq_action *h) rebalance_domains(this_cpu, idle); -#ifdef CONFIG_NO_HZ /* - * If this cpu is the owner for idle load balancing, then do the + * If this cpu has a pending nohz_balance_kick, then do the * balancing on behalf of the other idle cpus whose ticks are * stopped. */ - if (this_rq->idle_at_tick && - atomic_read(&nohz.load_balancer) == this_cpu) { - struct rq *rq; - int balance_cpu; - - for_each_cpu(balance_cpu, nohz.cpu_mask) { - if (balance_cpu == this_cpu) - continue; - - /* - * If this cpu gets work to do, stop the load balancing - * work being done for other cpus. Next load - * balancing owner will pick it up. - */ - if (need_resched()) - break; - - rebalance_domains(balance_cpu, CPU_IDLE); - - rq = cpu_rq(balance_cpu); - if (time_after(this_rq->next_balance, rq->next_balance)) - this_rq->next_balance = rq->next_balance; - } - } -#endif + nohz_idle_balance(this_cpu, idle); } static inline int on_null_domain(int cpu) @@ -3481,57 +3689,17 @@ static inline int on_null_domain(int cpu) /* * Trigger the SCHED_SOFTIRQ if it is time to do periodic load balancing. - * - * In case of CONFIG_NO_HZ, this is the place where we nominate a new - * idle load balancing owner or decide to stop the periodic load balancing, - * if the whole system is idle. */ static inline void trigger_load_balance(struct rq *rq, int cpu) { -#ifdef CONFIG_NO_HZ - /* - * If we were in the nohz mode recently and busy at the current - * scheduler tick, then check if we need to nominate new idle - * load balancer. - */ - if (rq->in_nohz_recently && !rq->idle_at_tick) { - rq->in_nohz_recently = 0; - - if (atomic_read(&nohz.load_balancer) == cpu) { - cpumask_clear_cpu(cpu, nohz.cpu_mask); - atomic_set(&nohz.load_balancer, -1); - } - - if (atomic_read(&nohz.load_balancer) == -1) { - int ilb = find_new_ilb(cpu); - - if (ilb < nr_cpu_ids) - resched_cpu(ilb); - } - } - - /* - * If this cpu is idle and doing idle load balancing for all the - * cpus with ticks stopped, is it time for that to stop? - */ - if (rq->idle_at_tick && atomic_read(&nohz.load_balancer) == cpu && - cpumask_weight(nohz.cpu_mask) == num_online_cpus()) { - resched_cpu(cpu); - return; - } - - /* - * If this cpu is idle and the idle load balancing is done by - * someone else, then no need raise the SCHED_SOFTIRQ - */ - if (rq->idle_at_tick && atomic_read(&nohz.load_balancer) != cpu && - cpumask_test_cpu(cpu, nohz.cpu_mask)) - return; -#endif /* Don't need to rebalance while attached to NULL domain */ if (time_after_eq(jiffies, rq->next_balance) && likely(!on_null_domain(cpu))) raise_softirq(SCHED_SOFTIRQ); +#ifdef CONFIG_NO_HZ + else if (nohz_kick_needed(rq, cpu) && likely(!on_null_domain(cpu))) + nohz_balancer_kick(cpu); +#endif } static void rq_online_fair(struct rq *rq) |