From 4d9842776a23e52ec4c60e0a79f5e1bbe91e463e Mon Sep 17 00:00:00 2001 From: Gregory Haskins Date: Mon, 29 Dec 2008 09:39:49 -0500 Subject: [PATCH 01/46] sched: cleanup inc/dec_rt_tasks Move some common definitions up to the function prologe to simplify the body logic. Signed-off-by: Gregory Haskins --- kernel/sched_rt.c | 40 ++++++++++++++++------------------------ 1 file changed, 16 insertions(+), 24 deletions(-) diff --git a/kernel/sched_rt.c b/kernel/sched_rt.c index 1bbd99014011..0a5277233452 100644 --- a/kernel/sched_rt.c +++ b/kernel/sched_rt.c @@ -550,30 +550,28 @@ static void update_curr_rt(struct rq *rq) static inline void inc_rt_tasks(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) { - WARN_ON(!rt_prio(rt_se_prio(rt_se))); + int prio = rt_se_prio(rt_se); +#ifdef CONFIG_SMP + struct rq *rq = rq_of_rt_rq(rt_rq); +#endif + + WARN_ON(!rt_prio(prio)); rt_rq->rt_nr_running++; #if defined CONFIG_SMP || defined CONFIG_RT_GROUP_SCHED - if (rt_se_prio(rt_se) < rt_rq->highest_prio) { -#ifdef CONFIG_SMP - struct rq *rq = rq_of_rt_rq(rt_rq); -#endif + if (prio < rt_rq->highest_prio) { - rt_rq->highest_prio = rt_se_prio(rt_se); + rt_rq->highest_prio = prio; #ifdef CONFIG_SMP if (rq->online) - cpupri_set(&rq->rd->cpupri, rq->cpu, - rt_se_prio(rt_se)); + cpupri_set(&rq->rd->cpupri, rq->cpu, prio); #endif } #endif #ifdef CONFIG_SMP - if (rt_se->nr_cpus_allowed > 1) { - struct rq *rq = rq_of_rt_rq(rt_rq); - + if (rt_se->nr_cpus_allowed > 1) rq->rt.rt_nr_migratory++; - } - update_rt_migration(rq_of_rt_rq(rt_rq)); + update_rt_migration(rq); #endif #ifdef CONFIG_RT_GROUP_SCHED if (rt_se_boosted(rt_se)) @@ -590,6 +588,7 @@ static inline void dec_rt_tasks(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) { #ifdef CONFIG_SMP + struct rq *rq = rq_of_rt_rq(rt_rq); int highest_prio = rt_rq->highest_prio; #endif @@ -611,20 +610,13 @@ void dec_rt_tasks(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) rt_rq->highest_prio = MAX_RT_PRIO; #endif #ifdef CONFIG_SMP - if (rt_se->nr_cpus_allowed > 1) { - struct rq *rq = rq_of_rt_rq(rt_rq); + if (rt_se->nr_cpus_allowed > 1) rq->rt.rt_nr_migratory--; - } - if (rt_rq->highest_prio != highest_prio) { - struct rq *rq = rq_of_rt_rq(rt_rq); + if (rq->online && rt_rq->highest_prio != highest_prio) + cpupri_set(&rq->rd->cpupri, rq->cpu, rt_rq->highest_prio); - if (rq->online) - cpupri_set(&rq->rd->cpupri, rq->cpu, - rt_rq->highest_prio); - } - - update_rt_migration(rq_of_rt_rq(rt_rq)); + update_rt_migration(rq); #endif /* CONFIG_SMP */ #ifdef CONFIG_RT_GROUP_SCHED if (rt_se_boosted(rt_se)) From e864c499d9e57805ae1f9e7ea404dd223759cd53 Mon Sep 17 00:00:00 2001 From: Gregory Haskins Date: Mon, 29 Dec 2008 09:39:49 -0500 Subject: [PATCH 02/46] sched: track the next-highest priority on each runqueue We will use this later in the series to reduce the amount of rq-lock contention during a pull operation Signed-off-by: Gregory Haskins --- kernel/sched.c | 8 +++-- kernel/sched_rt.c | 81 +++++++++++++++++++++++++++++++++++------------ 2 files changed, 67 insertions(+), 22 deletions(-) diff --git a/kernel/sched.c b/kernel/sched.c index 756d981d91a4..7729f9a45a8b 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -463,7 +463,10 @@ struct rt_rq { struct rt_prio_array active; unsigned long rt_nr_running; #if defined CONFIG_SMP || defined CONFIG_RT_GROUP_SCHED - int highest_prio; /* highest queued rt task prio */ + struct { + int curr; /* highest queued rt task prio */ + int next; /* next highest */ + } highest_prio; #endif #ifdef CONFIG_SMP unsigned long rt_nr_migratory; @@ -8169,7 +8172,8 @@ static void init_rt_rq(struct rt_rq *rt_rq, struct rq *rq) __set_bit(MAX_RT_PRIO, array->bitmap); #if defined CONFIG_SMP || defined CONFIG_RT_GROUP_SCHED - rt_rq->highest_prio = MAX_RT_PRIO; + rt_rq->highest_prio.curr = MAX_RT_PRIO; + rt_rq->highest_prio.next = MAX_RT_PRIO; #endif #ifdef CONFIG_SMP rt_rq->rt_nr_migratory = 0; diff --git a/kernel/sched_rt.c b/kernel/sched_rt.c index 0a5277233452..ad36d7232236 100644 --- a/kernel/sched_rt.c +++ b/kernel/sched_rt.c @@ -108,7 +108,7 @@ static void sched_rt_rq_enqueue(struct rt_rq *rt_rq) if (rt_rq->rt_nr_running) { if (rt_se && !on_rt_rq(rt_se)) enqueue_rt_entity(rt_se); - if (rt_rq->highest_prio < curr->prio) + if (rt_rq->highest_prio.curr < curr->prio) resched_task(curr); } } @@ -473,7 +473,7 @@ static inline int rt_se_prio(struct sched_rt_entity *rt_se) struct rt_rq *rt_rq = group_rt_rq(rt_se); if (rt_rq) - return rt_rq->highest_prio; + return rt_rq->highest_prio.curr; #endif return rt_task_of(rt_se)->prio; @@ -547,6 +547,21 @@ static void update_curr_rt(struct rq *rq) } } +#if defined CONFIG_SMP || defined CONFIG_RT_GROUP_SCHED + +static struct task_struct *pick_next_highest_task_rt(struct rq *rq, int cpu); + +static inline int next_prio(struct rq *rq) +{ + struct task_struct *next = pick_next_highest_task_rt(rq, rq->cpu); + + if (next && rt_prio(next->prio)) + return next->prio; + else + return MAX_RT_PRIO; +} +#endif + static inline void inc_rt_tasks(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) { @@ -558,14 +573,32 @@ void inc_rt_tasks(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) WARN_ON(!rt_prio(prio)); rt_rq->rt_nr_running++; #if defined CONFIG_SMP || defined CONFIG_RT_GROUP_SCHED - if (prio < rt_rq->highest_prio) { + if (prio < rt_rq->highest_prio.curr) { - rt_rq->highest_prio = prio; + /* + * If the new task is higher in priority than anything on the + * run-queue, we have a new high that must be published to + * the world. We also know that the previous high becomes + * our next-highest. + */ + rt_rq->highest_prio.next = rt_rq->highest_prio.curr; + rt_rq->highest_prio.curr = prio; #ifdef CONFIG_SMP if (rq->online) cpupri_set(&rq->rd->cpupri, rq->cpu, prio); #endif - } + } else if (prio == rt_rq->highest_prio.curr) + /* + * If the next task is equal in priority to the highest on + * the run-queue, then we implicitly know that the next highest + * task cannot be any lower than current + */ + rt_rq->highest_prio.next = prio; + else if (prio < rt_rq->highest_prio.next) + /* + * Otherwise, we need to recompute next-highest + */ + rt_rq->highest_prio.next = next_prio(rq); #endif #ifdef CONFIG_SMP if (rt_se->nr_cpus_allowed > 1) @@ -589,7 +622,7 @@ void dec_rt_tasks(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) { #ifdef CONFIG_SMP struct rq *rq = rq_of_rt_rq(rt_rq); - int highest_prio = rt_rq->highest_prio; + int highest_prio = rt_rq->highest_prio.curr; #endif WARN_ON(!rt_prio(rt_se_prio(rt_se))); @@ -597,24 +630,32 @@ void dec_rt_tasks(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) rt_rq->rt_nr_running--; #if defined CONFIG_SMP || defined CONFIG_RT_GROUP_SCHED if (rt_rq->rt_nr_running) { - struct rt_prio_array *array; + int prio = rt_se_prio(rt_se); - WARN_ON(rt_se_prio(rt_se) < rt_rq->highest_prio); - if (rt_se_prio(rt_se) == rt_rq->highest_prio) { - /* recalculate */ - array = &rt_rq->active; - rt_rq->highest_prio = + WARN_ON(prio < rt_rq->highest_prio.curr); + + /* + * This may have been our highest or next-highest priority + * task and therefore we may have some recomputation to do + */ + if (prio == rt_rq->highest_prio.curr) { + struct rt_prio_array *array = &rt_rq->active; + + rt_rq->highest_prio.curr = sched_find_first_bit(array->bitmap); - } /* otherwise leave rq->highest prio alone */ + } + + if (prio <= rt_rq->highest_prio.next) + rt_rq->highest_prio.next = next_prio(rq); } else - rt_rq->highest_prio = MAX_RT_PRIO; + rt_rq->highest_prio.curr = MAX_RT_PRIO; #endif #ifdef CONFIG_SMP if (rt_se->nr_cpus_allowed > 1) rq->rt.rt_nr_migratory--; - if (rq->online && rt_rq->highest_prio != highest_prio) - cpupri_set(&rq->rd->cpupri, rq->cpu, rt_rq->highest_prio); + if (rq->online && rt_rq->highest_prio.curr != highest_prio) + cpupri_set(&rq->rd->cpupri, rq->cpu, rt_rq->highest_prio.curr); update_rt_migration(rq); #endif /* CONFIG_SMP */ @@ -1064,7 +1105,7 @@ static struct rq *find_lock_lowest_rq(struct task_struct *task, struct rq *rq) } /* If this rq is still suitable use it. */ - if (lowest_rq->rt.highest_prio > task->prio) + if (lowest_rq->rt.highest_prio.curr > task->prio) break; /* try again */ @@ -1252,7 +1293,7 @@ static int pull_rt_task(struct rq *this_rq) static void pre_schedule_rt(struct rq *rq, struct task_struct *prev) { /* Try to pull RT tasks here if we lower this rq's prio */ - if (unlikely(rt_task(prev)) && rq->rt.highest_prio > prev->prio) + if (unlikely(rt_task(prev)) && rq->rt.highest_prio.curr > prev->prio) pull_rt_task(rq); } @@ -1338,7 +1379,7 @@ static void rq_online_rt(struct rq *rq) __enable_runtime(rq); - cpupri_set(&rq->rd->cpupri, rq->cpu, rq->rt.highest_prio); + cpupri_set(&rq->rd->cpupri, rq->cpu, rq->rt.highest_prio.curr); } /* Assumes rq->lock is held */ @@ -1429,7 +1470,7 @@ static void prio_changed_rt(struct rq *rq, struct task_struct *p, * can release the rq lock and p could migrate. * Only reschedule if p is still on the same runqueue. */ - if (p->prio > rq->rt.highest_prio && rq->curr == p) + if (p->prio > rq->rt.highest_prio.curr && rq->curr == p) resched_task(p); #else /* For UP simply resched on drop of prio */ From a8728944efe23417e38bf22063f06d9d8ee21d59 Mon Sep 17 00:00:00 2001 From: Gregory Haskins Date: Mon, 29 Dec 2008 09:39:49 -0500 Subject: [PATCH 03/46] sched: use highest_prio.curr for pull threshold highest_prio.curr is actually a more accurate way to keep track of the pull_rt_task() threshold since it is always up to date, even if the "next" task migrates during double_lock. Therefore, stop looking at the "next" task object and simply use the highest_prio.curr. Signed-off-by: Gregory Haskins --- kernel/sched_rt.c | 31 ++++++------------------------- 1 file changed, 6 insertions(+), 25 deletions(-) diff --git a/kernel/sched_rt.c b/kernel/sched_rt.c index ad36d7232236..f8fb3edadcaa 100644 --- a/kernel/sched_rt.c +++ b/kernel/sched_rt.c @@ -1207,14 +1207,12 @@ static void push_rt_tasks(struct rq *rq) static int pull_rt_task(struct rq *this_rq) { int this_cpu = this_rq->cpu, ret = 0, cpu; - struct task_struct *p, *next; + struct task_struct *p; struct rq *src_rq; if (likely(!rt_overloaded(this_rq))) return 0; - next = pick_next_task_rt(this_rq); - for_each_cpu(cpu, this_rq->rd->rto_mask) { if (this_cpu == cpu) continue; @@ -1223,17 +1221,9 @@ static int pull_rt_task(struct rq *this_rq) /* * We can potentially drop this_rq's lock in * double_lock_balance, and another CPU could - * steal our next task - hence we must cause - * the caller to recalculate the next task - * in that case: + * alter this_rq */ - if (double_lock_balance(this_rq, src_rq)) { - struct task_struct *old_next = next; - - next = pick_next_task_rt(this_rq); - if (next != old_next) - ret = 1; - } + double_lock_balance(this_rq, src_rq); /* * Are there still pullable RT tasks? @@ -1247,7 +1237,7 @@ static int pull_rt_task(struct rq *this_rq) * Do we have an RT task that preempts * the to-be-scheduled task? */ - if (p && (!next || (p->prio < next->prio))) { + if (p && (p->prio < this_rq->rt.highest_prio.curr)) { WARN_ON(p == src_rq->curr); WARN_ON(!p->se.on_rq); @@ -1257,12 +1247,9 @@ static int pull_rt_task(struct rq *this_rq) * This is just that p is wakeing up and hasn't * had a chance to schedule. We only pull * p if it is lower in priority than the - * current task on the run queue or - * this_rq next task is lower in prio than - * the current task on that rq. + * current task on the run queue */ - if (p->prio < src_rq->curr->prio || - (next && next->prio < src_rq->curr->prio)) + if (p->prio < src_rq->curr->prio) goto skip; ret = 1; @@ -1275,13 +1262,7 @@ static int pull_rt_task(struct rq *this_rq) * case there's an even higher prio task * in another runqueue. (low likelyhood * but possible) - * - * Update next so that we won't pick a task - * on another cpu with a priority lower (or equal) - * than the one we just picked. */ - next = p; - } skip: double_unlock_balance(this_rq, src_rq); From 74ab8e4f6412c0b2d730fe5de28dc21de8b92c01 Mon Sep 17 00:00:00 2001 From: Gregory Haskins Date: Mon, 29 Dec 2008 09:39:50 -0500 Subject: [PATCH 04/46] sched: use highest_prio.next to optimize pull operations We currently take the rq->lock for every cpu in an overload state during pull_rt_tasks(). However, we now have enough information via the highest_prio.[curr|next] fields to determine if there is any tasks of interest to warrant the overhead of the rq->lock, before we actually take it. So we use this information to reduce lock contention during the pull for the case where the source-rq doesnt have tasks that preempt the current task. Signed-off-by: Gregory Haskins --- kernel/sched_rt.c | 12 ++++++++++++ 1 file changed, 12 insertions(+) diff --git a/kernel/sched_rt.c b/kernel/sched_rt.c index f8fb3edadcaa..d047f288c411 100644 --- a/kernel/sched_rt.c +++ b/kernel/sched_rt.c @@ -1218,6 +1218,18 @@ static int pull_rt_task(struct rq *this_rq) continue; src_rq = cpu_rq(cpu); + + /* + * Don't bother taking the src_rq->lock if the next highest + * task is known to be lower-priority than our current task. + * This may look racy, but if this value is about to go + * logically higher, the src_rq will push this task away. + * And if its going logically lower, we do not care + */ + if (src_rq->rt.highest_prio.next >= + this_rq->rt.highest_prio.curr) + continue; + /* * We can potentially drop this_rq's lock in * double_lock_balance, and another CPU could From 777c2f389e463428fd7e2871051a84d7fe84b172 Mon Sep 17 00:00:00 2001 From: Gregory Haskins Date: Mon, 29 Dec 2008 09:39:50 -0500 Subject: [PATCH 05/46] sched: only try to push a task on wakeup if it is migratable There is no sense in wasting time trying to push a task away that cannot move anywhere else. We gain no benefit from trying to push other tasks at this point, so if the task being woken up is non migratable, just skip the whole operation. This reduces overhead in the wakeup path for certain tasks. Signed-off-by: Gregory Haskins --- kernel/sched_rt.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/kernel/sched_rt.c b/kernel/sched_rt.c index d047f288c411..8d33843cb2c4 100644 --- a/kernel/sched_rt.c +++ b/kernel/sched_rt.c @@ -1314,7 +1314,8 @@ static void task_wake_up_rt(struct rq *rq, struct task_struct *p) { if (!task_running(rq, p) && !test_tsk_need_resched(rq->curr) && - rq->rt.overloaded) + rq->rt.overloaded && + p->rt.nr_cpus_allowed > 1) push_rt_tasks(rq); } From 7e96fa5875d4a9be18d74d3ca7b90518d05bc426 Mon Sep 17 00:00:00 2001 From: Gregory Haskins Date: Mon, 29 Dec 2008 09:39:50 -0500 Subject: [PATCH 06/46] sched: pull only one task during NEWIDLE balancing to limit critical section git-id c4acb2c0669c5c5c9b28e9d02a34b5c67edf7092 attempted to limit newidle critical section length by stopping after at least one task was moved. Further investigation has shown that there are other paths nested further inside the algorithm which still remain that allow long latencies to occur with newidle balancing. This patch applies the same technique inside balance_tasks() to limit the duration of this optional balancing operation. Signed-off-by: Gregory Haskins CC: Nick Piggin --- kernel/sched.c | 18 +++++++++++++++++- 1 file changed, 17 insertions(+), 1 deletion(-) diff --git a/kernel/sched.c b/kernel/sched.c index 7729f9a45a8b..94d9a6c5ff94 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -2984,6 +2984,16 @@ next: pulled++; rem_load_move -= p->se.load.weight; +#ifdef CONFIG_PREEMPT + /* + * NEWIDLE balancing is a source of latency, so preemptible kernels + * will stop after the first task is pulled to minimize the critical + * section. + */ + if (idle == CPU_NEWLY_IDLE) + goto out; +#endif + /* * We only want to steal up to the prescribed amount of weighted load. */ @@ -3030,9 +3040,15 @@ static int move_tasks(struct rq *this_rq, int this_cpu, struct rq *busiest, sd, idle, all_pinned, &this_best_prio); class = class->next; +#ifdef CONFIG_PREEMPT + /* + * NEWIDLE balancing is a source of latency, so preemptible + * kernels will stop after the first task is pulled to minimize + * the critical section. + */ if (idle == CPU_NEWLY_IDLE && this_rq->nr_running) break; - +#endif } while (class && max_load_move > total_load_moved); return total_load_moved > 0; From 8f45e2b516201d1bf681e6026fa5276385def565 Mon Sep 17 00:00:00 2001 From: Gregory Haskins Date: Mon, 29 Dec 2008 09:39:51 -0500 Subject: [PATCH 07/46] sched: make double-lock-balance fair double_lock balance() currently favors logically lower cpus since they often do not have to release their own lock to acquire a second lock. The result is that logically higher cpus can get starved when there is a lot of pressure on the RQs. This can result in higher latencies on higher cpu-ids. This patch makes the algorithm more fair by forcing all paths to have to release both locks before acquiring them again. Since callsites to double_lock_balance already consider it a potential preemption/reschedule point, they have the proper logic to recheck for atomicity violations. Signed-off-by: Gregory Haskins --- kernel/sched.c | 51 +++++++++++++++++++++++++++++++++++++++++++------- 1 file changed, 44 insertions(+), 7 deletions(-) diff --git a/kernel/sched.c b/kernel/sched.c index 94d9a6c5ff94..8fca364f3593 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -1608,21 +1608,42 @@ static inline void update_shares_locked(struct rq *rq, struct sched_domain *sd) #endif +#ifdef CONFIG_PREEMPT + /* - * double_lock_balance - lock the busiest runqueue, this_rq is locked already. + * fair double_lock_balance: Safely acquires both rq->locks in a fair + * way at the expense of forcing extra atomic operations in all + * invocations. This assures that the double_lock is acquired using the + * same underlying policy as the spinlock_t on this architecture, which + * reduces latency compared to the unfair variant below. However, it + * also adds more overhead and therefore may reduce throughput. */ -static int double_lock_balance(struct rq *this_rq, struct rq *busiest) +static inline int _double_lock_balance(struct rq *this_rq, struct rq *busiest) + __releases(this_rq->lock) + __acquires(busiest->lock) + __acquires(this_rq->lock) +{ + spin_unlock(&this_rq->lock); + double_rq_lock(this_rq, busiest); + + return 1; +} + +#else +/* + * Unfair double_lock_balance: Optimizes throughput at the expense of + * latency by eliminating extra atomic operations when the locks are + * already in proper order on entry. This favors lower cpu-ids and will + * grant the double lock to lower cpus over higher ids under contention, + * regardless of entry order into the function. + */ +static int _double_lock_balance(struct rq *this_rq, struct rq *busiest) __releases(this_rq->lock) __acquires(busiest->lock) __acquires(this_rq->lock) { int ret = 0; - if (unlikely(!irqs_disabled())) { - /* printk() doesn't work good under rq->lock */ - spin_unlock(&this_rq->lock); - BUG_ON(1); - } if (unlikely(!spin_trylock(&busiest->lock))) { if (busiest < this_rq) { spin_unlock(&this_rq->lock); @@ -1635,6 +1656,22 @@ static int double_lock_balance(struct rq *this_rq, struct rq *busiest) return ret; } +#endif /* CONFIG_PREEMPT */ + +/* + * double_lock_balance - lock the busiest runqueue, this_rq is locked already. + */ +static int double_lock_balance(struct rq *this_rq, struct rq *busiest) +{ + if (unlikely(!irqs_disabled())) { + /* printk() doesn't work good under rq->lock */ + spin_unlock(&this_rq->lock); + BUG_ON(1); + } + + return _double_lock_balance(this_rq, busiest); +} + static inline void double_unlock_balance(struct rq *this_rq, struct rq *busiest) __releases(busiest->lock) { From 967fc04671feea4dbf780c9e55a0bc8fcf68a14e Mon Sep 17 00:00:00 2001 From: Gregory Haskins Date: Mon, 29 Dec 2008 09:39:52 -0500 Subject: [PATCH 08/46] sched: add sched_class->needs_post_schedule() member We currently run class->post_schedule() outside of the rq->lock, which means that we need to test for the need to post_schedule outside of the lock to avoid a forced reacquistion. This is currently not a problem as we only look at rq->rt.overloaded. However, we want to enhance this going forward to look at more state to reduce the need to post_schedule to a bare minimum set. Therefore, we introduce a new member-func called needs_post_schedule() which tests for the post_schedule condtion without actually performing the work. Therefore it is safe to call this function before the rq->lock is released, because we are guaranteed not to drop the lock at an intermediate point (such as what post_schedule() may do). We will use this later in the series [ rostedt: removed paranoid BUG_ON ] Signed-off-by: Gregory Haskins --- include/linux/sched.h | 1 + kernel/sched.c | 8 +++++++- kernel/sched_rt.c | 24 ++++++++++++++---------- 3 files changed, 22 insertions(+), 11 deletions(-) diff --git a/include/linux/sched.h b/include/linux/sched.h index e5f928a079e8..836a86c32a65 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -1012,6 +1012,7 @@ struct sched_class { struct rq *busiest, struct sched_domain *sd, enum cpu_idle_type idle); void (*pre_schedule) (struct rq *this_rq, struct task_struct *task); + int (*needs_post_schedule) (struct rq *this_rq); void (*post_schedule) (struct rq *this_rq); void (*task_wake_up) (struct rq *this_rq, struct task_struct *task); diff --git a/kernel/sched.c b/kernel/sched.c index 8fca364f3593..3acbad8991a2 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -2621,6 +2621,12 @@ static void finish_task_switch(struct rq *rq, struct task_struct *prev) { struct mm_struct *mm = rq->prev_mm; long prev_state; +#ifdef CONFIG_SMP + int post_schedule = 0; + + if (current->sched_class->needs_post_schedule) + post_schedule = current->sched_class->needs_post_schedule(rq); +#endif rq->prev_mm = NULL; @@ -2639,7 +2645,7 @@ static void finish_task_switch(struct rq *rq, struct task_struct *prev) finish_arch_switch(prev); finish_lock_switch(rq, prev); #ifdef CONFIG_SMP - if (current->sched_class->post_schedule) + if (post_schedule) current->sched_class->post_schedule(rq); #endif diff --git a/kernel/sched_rt.c b/kernel/sched_rt.c index 8d33843cb2c4..b0b6ea4ed674 100644 --- a/kernel/sched_rt.c +++ b/kernel/sched_rt.c @@ -1290,20 +1290,23 @@ static void pre_schedule_rt(struct rq *rq, struct task_struct *prev) pull_rt_task(rq); } +/* + * assumes rq->lock is held + */ +static int needs_post_schedule_rt(struct rq *rq) +{ + return rq->rt.overloaded ? 1 : 0; +} + static void post_schedule_rt(struct rq *rq) { /* - * If we have more than one rt_task queued, then - * see if we can push the other rt_tasks off to other CPUS. - * Note we may release the rq lock, and since - * the lock was owned by prev, we need to release it - * first via finish_lock_switch and then reaquire it here. + * This is only called if needs_post_schedule_rt() indicates that + * we need to push tasks away */ - if (unlikely(rq->rt.overloaded)) { - spin_lock_irq(&rq->lock); - push_rt_tasks(rq); - spin_unlock_irq(&rq->lock); - } + spin_lock_irq(&rq->lock); + push_rt_tasks(rq); + spin_unlock_irq(&rq->lock); } /* @@ -1557,6 +1560,7 @@ static const struct sched_class rt_sched_class = { .rq_online = rq_online_rt, .rq_offline = rq_offline_rt, .pre_schedule = pre_schedule_rt, + .needs_post_schedule = needs_post_schedule_rt, .post_schedule = post_schedule_rt, .task_wake_up = task_wake_up_rt, .switched_from = switched_from_rt, From 4075134e40804821f90866d7de56802e4dcecb1e Mon Sep 17 00:00:00 2001 From: Gregory Haskins Date: Mon, 29 Dec 2008 09:39:53 -0500 Subject: [PATCH 09/46] plist: fix PLIST_NODE_INIT to work with debug enabled It seems that PLIST_NODE_INIT breaks if used and DEBUG_PI_LIST is defined. Since there are no current users of PLIST_NODE_INIT, this has gone undetected. This patch fixes the build issue that enables the DEBUG_PI_LIST later in the series when we use it in init_task.h Signed-off-by: Gregory Haskins --- include/linux/plist.h | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) diff --git a/include/linux/plist.h b/include/linux/plist.h index 85de2f055874..45926d77d6ac 100644 --- a/include/linux/plist.h +++ b/include/linux/plist.h @@ -96,6 +96,10 @@ struct plist_node { # define PLIST_HEAD_LOCK_INIT(_lock) #endif +#define _PLIST_HEAD_INIT(head) \ + .prio_list = LIST_HEAD_INIT((head).prio_list), \ + .node_list = LIST_HEAD_INIT((head).node_list) + /** * PLIST_HEAD_INIT - static struct plist_head initializer * @head: struct plist_head variable name @@ -103,8 +107,7 @@ struct plist_node { */ #define PLIST_HEAD_INIT(head, _lock) \ { \ - .prio_list = LIST_HEAD_INIT((head).prio_list), \ - .node_list = LIST_HEAD_INIT((head).node_list), \ + _PLIST_HEAD_INIT(head), \ PLIST_HEAD_LOCK_INIT(&(_lock)) \ } @@ -116,7 +119,7 @@ struct plist_node { #define PLIST_NODE_INIT(node, __prio) \ { \ .prio = (__prio), \ - .plist = PLIST_HEAD_INIT((node).plist, NULL), \ + .plist = { _PLIST_HEAD_INIT((node).plist) }, \ } /** From 917b627d4d981dc614519d7b34ea31a976b14e12 Mon Sep 17 00:00:00 2001 From: Gregory Haskins Date: Mon, 29 Dec 2008 09:39:53 -0500 Subject: [PATCH 10/46] sched: create "pushable_tasks" list to limit pushing to one attempt The RT scheduler employs a "push/pull" design to actively balance tasks within the system (on a per disjoint cpuset basis). When a task is awoken, it is immediately determined if there are any lower priority cpus which should be preempted. This is opposed to the way normal SCHED_OTHER tasks behave, which will wait for a periodic rebalancing operation to occur before spreading out load. When a particular RQ has more than 1 active RT task, it is said to be in an "overloaded" state. Once this occurs, the system enters the active balancing mode, where it will try to push the task away, or persuade a different cpu to pull it over. The system will stay in this state until the system falls back below the <= 1 queued RT task per RQ. However, the current implementation suffers from a limitation in the push logic. Once overloaded, all tasks (other than current) on the RQ are analyzed on every push operation, even if it was previously unpushable (due to affinity, etc). Whats more, the operation stops at the first task that is unpushable and will not look at items lower in the queue. This causes two problems: 1) We can have the same tasks analyzed over and over again during each push, which extends out the fast path in the scheduler for no gain. Consider a RQ that has dozens of tasks that are bound to a core. Each one of those tasks will be encountered and skipped for each push operation while they are queued. 2) There may be lower-priority tasks under the unpushable task that could have been successfully pushed, but will never be considered until either the unpushable task is cleared, or a pull operation succeeds. The net result is a potential latency source for mid priority tasks. This patch aims to rectify these two conditions by introducing a new priority sorted list: "pushable_tasks". A task is added to the list each time a task is activated or preempted. It is removed from the list any time it is deactivated, made current, or fails to push. This works because a task only needs to be attempted to push once. After an initial failure to push, the other cpus will eventually try to pull the task when the conditions are proper. This also solves the problem that we don't completely analyze all tasks due to encountering an unpushable tasks. Now every task will have a push attempted (when appropriate). This reduces latency both by shorting the critical section of the rq->lock for certain workloads, and by making sure the algorithm considers all eligible tasks in the system. [ rostedt: added a couple more BUG_ONs ] Signed-off-by: Gregory Haskins Acked-by: Steven Rostedt --- include/linux/init_task.h | 1 + include/linux/sched.h | 1 + kernel/sched.c | 4 ++ kernel/sched_rt.c | 119 ++++++++++++++++++++++++++++++++------ 4 files changed, 107 insertions(+), 18 deletions(-) diff --git a/include/linux/init_task.h b/include/linux/init_task.h index 23fd8909b9e5..6851225f44a7 100644 --- a/include/linux/init_task.h +++ b/include/linux/init_task.h @@ -140,6 +140,7 @@ extern struct group_info init_groups; .nr_cpus_allowed = NR_CPUS, \ }, \ .tasks = LIST_HEAD_INIT(tsk.tasks), \ + .pushable_tasks = PLIST_NODE_INIT(tsk.pushable_tasks, MAX_PRIO), \ .ptraced = LIST_HEAD_INIT(tsk.ptraced), \ .ptrace_entry = LIST_HEAD_INIT(tsk.ptrace_entry), \ .real_parent = &tsk, \ diff --git a/include/linux/sched.h b/include/linux/sched.h index 836a86c32a65..440cabb2d432 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -1179,6 +1179,7 @@ struct task_struct { #endif struct list_head tasks; + struct plist_node pushable_tasks; struct mm_struct *mm, *active_mm; diff --git a/kernel/sched.c b/kernel/sched.c index 3acbad8991a2..24ab80c28765 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -471,6 +471,7 @@ struct rt_rq { #ifdef CONFIG_SMP unsigned long rt_nr_migratory; int overloaded; + struct plist_head pushable_tasks; #endif int rt_throttled; u64 rt_time; @@ -2481,6 +2482,8 @@ void sched_fork(struct task_struct *p, int clone_flags) /* Want to start with kernel preemption disabled. */ task_thread_info(p)->preempt_count = 1; #endif + plist_node_init(&p->pushable_tasks, MAX_PRIO); + put_cpu(); } @@ -8237,6 +8240,7 @@ static void init_rt_rq(struct rt_rq *rt_rq, struct rq *rq) #ifdef CONFIG_SMP rt_rq->rt_nr_migratory = 0; rt_rq->overloaded = 0; + plist_head_init(&rq->rt.pushable_tasks, &rq->lock); #endif rt_rq->rt_time = 0; diff --git a/kernel/sched_rt.c b/kernel/sched_rt.c index b0b6ea4ed674..fe9da6084c87 100644 --- a/kernel/sched_rt.c +++ b/kernel/sched_rt.c @@ -49,6 +49,24 @@ static void update_rt_migration(struct rq *rq) rq->rt.overloaded = 0; } } + +static void enqueue_pushable_task(struct rq *rq, struct task_struct *p) +{ + plist_del(&p->pushable_tasks, &rq->rt.pushable_tasks); + plist_node_init(&p->pushable_tasks, p->prio); + plist_add(&p->pushable_tasks, &rq->rt.pushable_tasks); +} + +static void dequeue_pushable_task(struct rq *rq, struct task_struct *p) +{ + plist_del(&p->pushable_tasks, &rq->rt.pushable_tasks); +} + +#else + +#define enqueue_pushable_task(rq, p) do { } while (0) +#define dequeue_pushable_task(rq, p) do { } while (0) + #endif /* CONFIG_SMP */ static inline struct task_struct *rt_task_of(struct sched_rt_entity *rt_se) @@ -751,6 +769,9 @@ static void enqueue_task_rt(struct rq *rq, struct task_struct *p, int wakeup) enqueue_rt_entity(rt_se); + if (!task_current(rq, p) && p->rt.nr_cpus_allowed > 1) + enqueue_pushable_task(rq, p); + inc_cpu_load(rq, p->se.load.weight); } @@ -761,6 +782,8 @@ static void dequeue_task_rt(struct rq *rq, struct task_struct *p, int sleep) update_curr_rt(rq); dequeue_rt_entity(rt_se); + dequeue_pushable_task(rq, p); + dec_cpu_load(rq, p->se.load.weight); } @@ -911,7 +934,7 @@ static struct sched_rt_entity *pick_next_rt_entity(struct rq *rq, return next; } -static struct task_struct *pick_next_task_rt(struct rq *rq) +static struct task_struct *_pick_next_task_rt(struct rq *rq) { struct sched_rt_entity *rt_se; struct task_struct *p; @@ -933,6 +956,18 @@ static struct task_struct *pick_next_task_rt(struct rq *rq) p = rt_task_of(rt_se); p->se.exec_start = rq->clock; + + return p; +} + +static struct task_struct *pick_next_task_rt(struct rq *rq) +{ + struct task_struct *p = _pick_next_task_rt(rq); + + /* The running task is never eligible for pushing */ + if (p) + dequeue_pushable_task(rq, p); + return p; } @@ -940,6 +975,13 @@ static void put_prev_task_rt(struct rq *rq, struct task_struct *p) { update_curr_rt(rq); p->se.exec_start = 0; + + /* + * The previous task needs to be made eligible for pushing + * if it is still active + */ + if (p->se.on_rq && p->rt.nr_cpus_allowed > 1) + enqueue_pushable_task(rq, p); } #ifdef CONFIG_SMP @@ -1116,6 +1158,31 @@ static struct rq *find_lock_lowest_rq(struct task_struct *task, struct rq *rq) return lowest_rq; } +static inline int has_pushable_tasks(struct rq *rq) +{ + return !plist_head_empty(&rq->rt.pushable_tasks); +} + +static struct task_struct *pick_next_pushable_task(struct rq *rq) +{ + struct task_struct *p; + + if (!has_pushable_tasks(rq)) + return NULL; + + p = plist_first_entry(&rq->rt.pushable_tasks, + struct task_struct, pushable_tasks); + + BUG_ON(rq->cpu != task_cpu(p)); + BUG_ON(task_current(rq, p)); + BUG_ON(p->rt.nr_cpus_allowed <= 1); + + BUG_ON(!p->se.on_rq); + BUG_ON(!rt_task(p)); + + return p; +} + /* * If the current CPU has more than one RT task, see if the non * running task can migrate over to a CPU that is running a task @@ -1125,13 +1192,12 @@ static int push_rt_task(struct rq *rq) { struct task_struct *next_task; struct rq *lowest_rq; - int ret = 0; int paranoid = RT_MAX_TRIES; if (!rq->rt.overloaded) return 0; - next_task = pick_next_highest_task_rt(rq, -1); + next_task = pick_next_pushable_task(rq); if (!next_task) return 0; @@ -1163,12 +1229,19 @@ static int push_rt_task(struct rq *rq) * so it is possible that next_task has changed. * If it has, then try again. */ - task = pick_next_highest_task_rt(rq, -1); + task = pick_next_pushable_task(rq); if (unlikely(task != next_task) && task && paranoid--) { put_task_struct(next_task); next_task = task; goto retry; } + + /* + * Once we have failed to push this task, we will not + * try again, since the other cpus will pull from us + * when they are ready + */ + dequeue_pushable_task(rq, next_task); goto out; } @@ -1180,23 +1253,12 @@ static int push_rt_task(struct rq *rq) double_unlock_balance(rq, lowest_rq); - ret = 1; out: put_task_struct(next_task); - return ret; + return 1; } -/* - * TODO: Currently we just use the second highest prio task on - * the queue, and stop when it can't migrate (or there's - * no more RT tasks). There may be a case where a lower - * priority RT task has a different affinity than the - * higher RT task. In this case the lower RT task could - * possibly be able to migrate where as the higher priority - * RT task could not. We currently ignore this issue. - * Enhancements are welcome! - */ static void push_rt_tasks(struct rq *rq) { /* push_rt_task will return true if it moved an RT */ @@ -1295,7 +1357,7 @@ static void pre_schedule_rt(struct rq *rq, struct task_struct *prev) */ static int needs_post_schedule_rt(struct rq *rq) { - return rq->rt.overloaded ? 1 : 0; + return has_pushable_tasks(rq); } static void post_schedule_rt(struct rq *rq) @@ -1317,7 +1379,7 @@ static void task_wake_up_rt(struct rq *rq, struct task_struct *p) { if (!task_running(rq, p) && !test_tsk_need_resched(rq->curr) && - rq->rt.overloaded && + has_pushable_tasks(rq) && p->rt.nr_cpus_allowed > 1) push_rt_tasks(rq); } @@ -1354,6 +1416,24 @@ static void set_cpus_allowed_rt(struct task_struct *p, if (p->se.on_rq && (weight != p->rt.nr_cpus_allowed)) { struct rq *rq = task_rq(p); + if (!task_current(rq, p)) { + /* + * Make sure we dequeue this task from the pushable list + * before going further. It will either remain off of + * the list because we are no longer pushable, or it + * will be requeued. + */ + if (p->rt.nr_cpus_allowed > 1) + dequeue_pushable_task(rq, p); + + /* + * Requeue if our weight is changing and still > 1 + */ + if (weight > 1) + enqueue_pushable_task(rq, p); + + } + if ((p->rt.nr_cpus_allowed <= 1) && (weight > 1)) { rq->rt.rt_nr_migratory++; } else if ((p->rt.nr_cpus_allowed > 1) && (weight <= 1)) { @@ -1538,6 +1618,9 @@ static void set_curr_task_rt(struct rq *rq) struct task_struct *p = rq->curr; p->se.exec_start = rq->clock; + + /* The running task is never eligible for pushing */ + dequeue_pushable_task(rq, p); } static const struct sched_class rt_sched_class = { From 1563513d34ed4b12ef32bc2adde4a53ce05701a1 Mon Sep 17 00:00:00 2001 From: Gregory Haskins Date: Mon, 29 Dec 2008 09:39:53 -0500 Subject: [PATCH 11/46] RT: fix push_rt_task() to handle dequeue_pushable properly A panic was discovered by Chirag Jog where a BUG_ON sanity check in the new "pushable_task" logic would trigger a panic under certain circumstances: http://lkml.org/lkml/2008/9/25/189 Gilles Carry discovered that the root cause was attributed to the pushable_tasks list getting corrupted in the push_rt_task logic. This was the result of a dropped rq lock in double_lock_balance allowing a task in the process of being pushed to potentially migrate away, and thus corrupt the pushable_tasks() list. I traced back the problem as introduced by the pushable_tasks patch that went in recently. There is a "retry" path in push_rt_task() that actually had a compound conditional to decide whether to retry or exit. I missed the meaning behind the rationale for the virtual "if(!task) goto out;" portion of the compound statement and thus did not handle it properly. The new pushable_tasks logic actually creates three distinct conditions: 1) an untouched and unpushable task should be dequeued 2) a migrated task where more pushable tasks remain should be retried 3) a migrated task where no more pushable tasks exist should exit The original logic mushed (1) and (3) together, resulting in the system dequeuing a migrated task (against an unlocked foreign run-queue nonetheless). To fix this, we get rid of the notion of "paranoid" and we support the three unique conditions properly. The paranoid feature is no longer relevant with the new pushable logic (since pushable naturally limits the loop) anyway, so lets just remove it. Reported-By: Chirag Jog Found-by: Gilles Carry Signed-off-by: Gregory Haskins --- kernel/sched_rt.c | 34 ++++++++++++++++++++++------------ 1 file changed, 22 insertions(+), 12 deletions(-) diff --git a/kernel/sched_rt.c b/kernel/sched_rt.c index fe9da6084c87..64a8f0aa117b 100644 --- a/kernel/sched_rt.c +++ b/kernel/sched_rt.c @@ -1192,7 +1192,6 @@ static int push_rt_task(struct rq *rq) { struct task_struct *next_task; struct rq *lowest_rq; - int paranoid = RT_MAX_TRIES; if (!rq->rt.overloaded) return 0; @@ -1226,23 +1225,34 @@ static int push_rt_task(struct rq *rq) struct task_struct *task; /* * find lock_lowest_rq releases rq->lock - * so it is possible that next_task has changed. - * If it has, then try again. + * so it is possible that next_task has migrated. + * + * We need to make sure that the task is still on the same + * run-queue and is also still the next task eligible for + * pushing. */ task = pick_next_pushable_task(rq); - if (unlikely(task != next_task) && task && paranoid--) { - put_task_struct(next_task); - next_task = task; - goto retry; + if (task_cpu(next_task) == rq->cpu && task == next_task) { + /* + * If we get here, the task hasnt moved at all, but + * it has failed to push. We will not try again, + * since the other cpus will pull from us when they + * are ready. + */ + dequeue_pushable_task(rq, next_task); + goto out; } + if (!task) + /* No more tasks, just exit */ + goto out; + /* - * Once we have failed to push this task, we will not - * try again, since the other cpus will pull from us - * when they are ready + * Something has shifted, try again. */ - dequeue_pushable_task(rq, next_task); - goto out; + put_task_struct(next_task); + next_task = task; + goto retry; } deactivate_task(rq, next_task, 0); From b07430ac37103218b5c1e542490a1b98e6deb3d6 Mon Sep 17 00:00:00 2001 From: Gregory Haskins Date: Wed, 14 Jan 2009 08:55:39 -0500 Subject: [PATCH 12/46] sched: de CPP-ify the scheduler code Signed-off-by: Gregory Haskins --- kernel/sched_rt.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) diff --git a/kernel/sched_rt.c b/kernel/sched_rt.c index 18c7b5b3158a..4eda5f795f04 100644 --- a/kernel/sched_rt.c +++ b/kernel/sched_rt.c @@ -64,8 +64,10 @@ static void dequeue_pushable_task(struct rq *rq, struct task_struct *p) #else -#define enqueue_pushable_task(rq, p) do { } while (0) -#define dequeue_pushable_task(rq, p) do { } while (0) +static inline +void enqueue_pushable_task(struct rq *rq, struct task_struct *p) {} +static inline +void dequeue_pushable_task(struct rq *rq, struct task_struct *p) {} #endif /* CONFIG_SMP */ From 398a153b16b09a68739928d4502455db9725ac86 Mon Sep 17 00:00:00 2001 From: Gregory Haskins Date: Wed, 14 Jan 2009 09:10:04 -0500 Subject: [PATCH 13/46] sched: fix build error in kernel/sched_rt.c when RT_GROUP_SCHED && !SMP Ingo found a build error in the scheduler when RT_GROUP_SCHED was enabled, but SMP was not. This patch rearranges the code such that it is a little more streamlined and compiles under all permutations of SMP, UP and RT_GROUP_SCHED. It was boot tested on my 4-way x86_64 and it still passes preempt-test. Signed-off-by: Gregory Haskins --- kernel/sched.c | 4 + kernel/sched_rt.c | 268 +++++++++++++++++++++++++++++----------------- 2 files changed, 176 insertions(+), 96 deletions(-) diff --git a/kernel/sched.c b/kernel/sched.c index dd1a1466c1e6..2b703f1fac3a 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -466,7 +466,9 @@ struct rt_rq { #if defined CONFIG_SMP || defined CONFIG_RT_GROUP_SCHED struct { int curr; /* highest queued rt task prio */ +#ifdef CONFIG_SMP int next; /* next highest */ +#endif } highest_prio; #endif #ifdef CONFIG_SMP @@ -8267,8 +8269,10 @@ static void init_rt_rq(struct rt_rq *rt_rq, struct rq *rq) #if defined CONFIG_SMP || defined CONFIG_RT_GROUP_SCHED rt_rq->highest_prio.curr = MAX_RT_PRIO; +#ifdef CONFIG_SMP rt_rq->highest_prio.next = MAX_RT_PRIO; #endif +#endif #ifdef CONFIG_SMP rt_rq->rt_nr_migratory = 0; rt_rq->overloaded = 0; diff --git a/kernel/sched_rt.c b/kernel/sched_rt.c index 4eda5f795f04..4230b15fe90e 100644 --- a/kernel/sched_rt.c +++ b/kernel/sched_rt.c @@ -3,6 +3,40 @@ * policies) */ +static inline struct task_struct *rt_task_of(struct sched_rt_entity *rt_se) +{ + return container_of(rt_se, struct task_struct, rt); +} + +#ifdef CONFIG_RT_GROUP_SCHED + +static inline struct rq *rq_of_rt_rq(struct rt_rq *rt_rq) +{ + return rt_rq->rq; +} + +static inline struct rt_rq *rt_rq_of_se(struct sched_rt_entity *rt_se) +{ + return rt_se->rt_rq; +} + +#else /* CONFIG_RT_GROUP_SCHED */ + +static inline struct rq *rq_of_rt_rq(struct rt_rq *rt_rq) +{ + return container_of(rt_rq, struct rq, rt); +} + +static inline struct rt_rq *rt_rq_of_se(struct sched_rt_entity *rt_se) +{ + struct task_struct *p = rt_task_of(rt_se); + struct rq *rq = task_rq(p); + + return &rq->rt; +} + +#endif /* CONFIG_RT_GROUP_SCHED */ + #ifdef CONFIG_SMP static inline int rt_overloaded(struct rq *rq) @@ -37,19 +71,35 @@ static inline void rt_clear_overload(struct rq *rq) cpumask_clear_cpu(rq->cpu, rq->rd->rto_mask); } -static void update_rt_migration(struct rq *rq) +static void update_rt_migration(struct rt_rq *rt_rq) { - if (rq->rt.rt_nr_migratory && (rq->rt.rt_nr_running > 1)) { - if (!rq->rt.overloaded) { - rt_set_overload(rq); - rq->rt.overloaded = 1; + if (rt_rq->rt_nr_migratory && (rt_rq->rt_nr_running > 1)) { + if (!rt_rq->overloaded) { + rt_set_overload(rq_of_rt_rq(rt_rq)); + rt_rq->overloaded = 1; } - } else if (rq->rt.overloaded) { - rt_clear_overload(rq); - rq->rt.overloaded = 0; + } else if (rt_rq->overloaded) { + rt_clear_overload(rq_of_rt_rq(rt_rq)); + rt_rq->overloaded = 0; } } +static void inc_rt_migration(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) +{ + if (rt_se->nr_cpus_allowed > 1) + rt_rq->rt_nr_migratory++; + + update_rt_migration(rt_rq); +} + +static void dec_rt_migration(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) +{ + if (rt_se->nr_cpus_allowed > 1) + rt_rq->rt_nr_migratory--; + + update_rt_migration(rt_rq); +} + static void enqueue_pushable_task(struct rq *rq, struct task_struct *p) { plist_del(&p->pushable_tasks, &rq->rt.pushable_tasks); @@ -68,14 +118,13 @@ static inline void enqueue_pushable_task(struct rq *rq, struct task_struct *p) {} static inline void dequeue_pushable_task(struct rq *rq, struct task_struct *p) {} +static inline +void inc_rt_migration(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) {} +static inline +void dec_rt_migration(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) {} #endif /* CONFIG_SMP */ -static inline struct task_struct *rt_task_of(struct sched_rt_entity *rt_se) -{ - return container_of(rt_se, struct task_struct, rt); -} - static inline int on_rt_rq(struct sched_rt_entity *rt_se) { return !list_empty(&rt_se->run_list); @@ -99,16 +148,6 @@ static inline u64 sched_rt_period(struct rt_rq *rt_rq) #define for_each_leaf_rt_rq(rt_rq, rq) \ list_for_each_entry_rcu(rt_rq, &rq->leaf_rt_rq_list, leaf_rt_rq_list) -static inline struct rq *rq_of_rt_rq(struct rt_rq *rt_rq) -{ - return rt_rq->rq; -} - -static inline struct rt_rq *rt_rq_of_se(struct sched_rt_entity *rt_se) -{ - return rt_se->rt_rq; -} - #define for_each_sched_rt_entity(rt_se) \ for (; rt_se; rt_se = rt_se->parent) @@ -196,19 +235,6 @@ static inline u64 sched_rt_period(struct rt_rq *rt_rq) #define for_each_leaf_rt_rq(rt_rq, rq) \ for (rt_rq = &rq->rt; rt_rq; rt_rq = NULL) -static inline struct rq *rq_of_rt_rq(struct rt_rq *rt_rq) -{ - return container_of(rt_rq, struct rq, rt); -} - -static inline struct rt_rq *rt_rq_of_se(struct sched_rt_entity *rt_se) -{ - struct task_struct *p = rt_task_of(rt_se); - struct rq *rq = task_rq(p); - - return &rq->rt; -} - #define for_each_sched_rt_entity(rt_se) \ for (; rt_se; rt_se = NULL) @@ -567,7 +593,7 @@ static void update_curr_rt(struct rq *rq) } } -#if defined CONFIG_SMP || defined CONFIG_RT_GROUP_SCHED +#if defined CONFIG_SMP static struct task_struct *pick_next_highest_task_rt(struct rq *rq, int cpu); @@ -580,33 +606,24 @@ static inline int next_prio(struct rq *rq) else return MAX_RT_PRIO; } -#endif -static inline -void inc_rt_tasks(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) +static void +inc_rt_prio_smp(struct rt_rq *rt_rq, int prio, int prev_prio) { - int prio = rt_se_prio(rt_se); -#ifdef CONFIG_SMP struct rq *rq = rq_of_rt_rq(rt_rq); -#endif - WARN_ON(!rt_prio(prio)); - rt_rq->rt_nr_running++; -#if defined CONFIG_SMP || defined CONFIG_RT_GROUP_SCHED - if (prio < rt_rq->highest_prio.curr) { + if (prio < prev_prio) { /* * If the new task is higher in priority than anything on the - * run-queue, we have a new high that must be published to - * the world. We also know that the previous high becomes - * our next-highest. + * run-queue, we know that the previous high becomes our + * next-highest. */ - rt_rq->highest_prio.next = rt_rq->highest_prio.curr; - rt_rq->highest_prio.curr = prio; -#ifdef CONFIG_SMP + rt_rq->highest_prio.next = prev_prio; + if (rq->online) cpupri_set(&rq->rd->cpupri, rq->cpu, prio); -#endif + } else if (prio == rt_rq->highest_prio.curr) /* * If the next task is equal in priority to the highest on @@ -619,72 +636,131 @@ void inc_rt_tasks(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) * Otherwise, we need to recompute next-highest */ rt_rq->highest_prio.next = next_prio(rq); -#endif -#ifdef CONFIG_SMP - if (rt_se->nr_cpus_allowed > 1) - rq->rt.rt_nr_migratory++; - - update_rt_migration(rq); -#endif -#ifdef CONFIG_RT_GROUP_SCHED - if (rt_se_boosted(rt_se)) - rt_rq->rt_nr_boosted++; - - if (rt_rq->tg) - start_rt_bandwidth(&rt_rq->tg->rt_bandwidth); -#else - start_rt_bandwidth(&def_rt_bandwidth); -#endif } -static inline -void dec_rt_tasks(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) +static void +dec_rt_prio_smp(struct rt_rq *rt_rq, int prio, int prev_prio) { -#ifdef CONFIG_SMP struct rq *rq = rq_of_rt_rq(rt_rq); - int highest_prio = rt_rq->highest_prio.curr; -#endif - WARN_ON(!rt_prio(rt_se_prio(rt_se))); - WARN_ON(!rt_rq->rt_nr_running); - rt_rq->rt_nr_running--; + if (rt_rq->rt_nr_running && (prio <= rt_rq->highest_prio.next)) + rt_rq->highest_prio.next = next_prio(rq); + + if (rq->online && rt_rq->highest_prio.curr != prev_prio) + cpupri_set(&rq->rd->cpupri, rq->cpu, rt_rq->highest_prio.curr); +} + +#else /* CONFIG_SMP */ + +static inline +void inc_rt_prio_smp(struct rt_rq *rt_rq, int prio, int prev_prio) {} +static inline +void dec_rt_prio_smp(struct rt_rq *rt_rq, int prio, int prev_prio) {} + +#endif /* CONFIG_SMP */ + #if defined CONFIG_SMP || defined CONFIG_RT_GROUP_SCHED - if (rt_rq->rt_nr_running) { - int prio = rt_se_prio(rt_se); +static void +inc_rt_prio(struct rt_rq *rt_rq, int prio) +{ + int prev_prio = rt_rq->highest_prio.curr; - WARN_ON(prio < rt_rq->highest_prio.curr); + if (prio < prev_prio) + rt_rq->highest_prio.curr = prio; + + inc_rt_prio_smp(rt_rq, prio, prev_prio); +} + +static void +dec_rt_prio(struct rt_rq *rt_rq, int prio) +{ + int prev_prio = rt_rq->highest_prio.curr; + + if (rt_rq->rt_nr_running) { + + WARN_ON(prio < prev_prio); /* - * This may have been our highest or next-highest priority - * task and therefore we may have some recomputation to do + * This may have been our highest task, and therefore + * we may have some recomputation to do */ - if (prio == rt_rq->highest_prio.curr) { + if (prio == prev_prio) { struct rt_prio_array *array = &rt_rq->active; rt_rq->highest_prio.curr = sched_find_first_bit(array->bitmap); } - if (prio <= rt_rq->highest_prio.next) - rt_rq->highest_prio.next = next_prio(rq); } else rt_rq->highest_prio.curr = MAX_RT_PRIO; -#endif -#ifdef CONFIG_SMP - if (rt_se->nr_cpus_allowed > 1) - rq->rt.rt_nr_migratory--; - if (rq->online && rt_rq->highest_prio.curr != highest_prio) - cpupri_set(&rq->rd->cpupri, rq->cpu, rt_rq->highest_prio.curr); + dec_rt_prio_smp(rt_rq, prio, prev_prio); +} + +#else + +static inline void inc_rt_prio(struct rt_rq *rt_rq, int prio) {} +static inline void dec_rt_prio(struct rt_rq *rt_rq, int prio) {} + +#endif /* CONFIG_SMP || CONFIG_RT_GROUP_SCHED */ - update_rt_migration(rq); -#endif /* CONFIG_SMP */ #ifdef CONFIG_RT_GROUP_SCHED + +static void +inc_rt_group(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) +{ + if (rt_se_boosted(rt_se)) + rt_rq->rt_nr_boosted++; + + if (rt_rq->tg) + start_rt_bandwidth(&rt_rq->tg->rt_bandwidth); +} + +static void +dec_rt_group(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) +{ if (rt_se_boosted(rt_se)) rt_rq->rt_nr_boosted--; WARN_ON(!rt_rq->rt_nr_running && rt_rq->rt_nr_boosted); -#endif +} + +#else /* CONFIG_RT_GROUP_SCHED */ + +static void +inc_rt_group(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) +{ + start_rt_bandwidth(&def_rt_bandwidth); +} + +static inline +void dec_rt_group(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) {} + +#endif /* CONFIG_RT_GROUP_SCHED */ + +static inline +void inc_rt_tasks(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) +{ + int prio = rt_se_prio(rt_se); + + WARN_ON(!rt_prio(prio)); + rt_rq->rt_nr_running++; + + inc_rt_prio(rt_rq, prio); + inc_rt_migration(rt_se, rt_rq); + inc_rt_group(rt_se, rt_rq); +} + +static inline +void dec_rt_tasks(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) +{ + WARN_ON(!rt_prio(rt_se_prio(rt_se))); + WARN_ON(!rt_rq->rt_nr_running); + rt_rq->rt_nr_running--; + + dec_rt_prio(rt_rq, rt_se_prio(rt_se)); + dec_rt_migration(rt_se, rt_rq); + dec_rt_group(rt_se, rt_rq); } static void __enqueue_rt_entity(struct sched_rt_entity *rt_se) @@ -1453,7 +1529,7 @@ static void set_cpus_allowed_rt(struct task_struct *p, rq->rt.rt_nr_migratory--; } - update_rt_migration(rq); + update_rt_migration(&rq->rt); } cpumask_copy(&p->cpus_allowed, new_mask); From 831451ac4e44d3a20b581ce726ef1d1144373f7d Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Wed, 14 Jan 2009 12:39:18 +0100 Subject: [PATCH 14/46] sched: introduce avg_wakeup Introduce a new avg_wakeup statistic. avg_wakeup is a measure of how frequently a task wakes up other tasks, it represents the average time between wakeups, with a limit of avg_runtime for when it doesn't wake up anybody. Signed-off-by: Peter Zijlstra Signed-off-by: Mike Galbraith Signed-off-by: Ingo Molnar --- include/linux/sched.h | 3 +++ kernel/sched.c | 36 ++++++++++++++++++++++++++++++------ kernel/sched_debug.c | 1 + 3 files changed, 34 insertions(+), 6 deletions(-) diff --git a/include/linux/sched.h b/include/linux/sched.h index 4cae9b81a1f8..daf4e07bc978 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -1046,6 +1046,9 @@ struct sched_entity { u64 exec_max; u64 slice_max; + u64 start_runtime; + u64 avg_wakeup; + u64 nr_migrations; u64 nr_migrations_cold; u64 nr_failed_migrations_affine; diff --git a/kernel/sched.c b/kernel/sched.c index 8be2c13b50d0..86f5a063f0b9 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -1705,6 +1705,9 @@ static void update_avg(u64 *avg, u64 sample) static void enqueue_task(struct rq *rq, struct task_struct *p, int wakeup) { + if (wakeup) + p->se.start_runtime = p->se.sum_exec_runtime; + sched_info_queued(p); p->sched_class->enqueue_task(rq, p, wakeup); p->se.on_rq = 1; @@ -1712,10 +1715,15 @@ static void enqueue_task(struct rq *rq, struct task_struct *p, int wakeup) static void dequeue_task(struct rq *rq, struct task_struct *p, int sleep) { - if (sleep && p->se.last_wakeup) { - update_avg(&p->se.avg_overlap, - p->se.sum_exec_runtime - p->se.last_wakeup); - p->se.last_wakeup = 0; + if (sleep) { + if (p->se.last_wakeup) { + update_avg(&p->se.avg_overlap, + p->se.sum_exec_runtime - p->se.last_wakeup); + p->se.last_wakeup = 0; + } else { + update_avg(&p->se.avg_wakeup, + sysctl_sched_wakeup_granularity); + } } sched_info_dequeued(p); @@ -2345,6 +2353,22 @@ out_activate: activate_task(rq, p, 1); success = 1; + /* + * Only attribute actual wakeups done by this task. + */ + if (!in_interrupt()) { + struct sched_entity *se = ¤t->se; + u64 sample = se->sum_exec_runtime; + + if (se->last_wakeup) + sample -= se->last_wakeup; + else + sample -= se->start_runtime; + update_avg(&se->avg_wakeup, sample); + + se->last_wakeup = se->sum_exec_runtime; + } + out_running: trace_sched_wakeup(rq, p, success); check_preempt_curr(rq, p, sync); @@ -2355,8 +2379,6 @@ out_running: p->sched_class->task_wake_up(rq, p); #endif out: - current->se.last_wakeup = current->se.sum_exec_runtime; - task_rq_unlock(rq, &flags); return success; @@ -2386,6 +2408,8 @@ static void __sched_fork(struct task_struct *p) p->se.prev_sum_exec_runtime = 0; p->se.last_wakeup = 0; p->se.avg_overlap = 0; + p->se.start_runtime = 0; + p->se.avg_wakeup = sysctl_sched_wakeup_granularity; #ifdef CONFIG_SCHEDSTATS p->se.wait_start = 0; diff --git a/kernel/sched_debug.c b/kernel/sched_debug.c index 16eeba4e4169..2b1260f0e800 100644 --- a/kernel/sched_debug.c +++ b/kernel/sched_debug.c @@ -397,6 +397,7 @@ void proc_sched_show_task(struct task_struct *p, struct seq_file *m) PN(se.vruntime); PN(se.sum_exec_runtime); PN(se.avg_overlap); + PN(se.avg_wakeup); nr_switches = p->nvcsw + p->nivcsw; From e52fb7c097238d34f4d8e2a596f8a3f85b0c0565 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Wed, 14 Jan 2009 12:39:19 +0100 Subject: [PATCH 15/46] sched: prefer wakers Prefer tasks that wake other tasks to preempt quickly. This improves performance because more work is available sooner. The workload that prompted this patch was a kernel build over NFS4 (for some curious and not understood reason we had to revert commit: 18de9735300756e3ca9c361ef58409d8561dfe0d to make any progress at all) Without this patch a make -j8 bzImage (of x86-64 defconfig) would take 3m30-ish, with this patch we're down to 2m50-ish. psql-sysbench/mysql-sysbench show a slight improvement in peak performance as well, tbench and vmark seemed to not care. It is possible to improve upon the build time (to 2m20-ish) but that seriously destroys other benchmarks (just shows that there's more room for tinkering). Much thanks to Mike who put in a lot of effort to benchmark things and proved a worthy opponent with a competing patch. Signed-off-by: Peter Zijlstra Signed-off-by: Mike Galbraith Signed-off-by: Ingo Molnar --- kernel/sched_fair.c | 59 ++++++++++++++++++++++++++++++++++++----- kernel/sched_features.h | 3 ++- 2 files changed, 55 insertions(+), 7 deletions(-) diff --git a/kernel/sched_fair.c b/kernel/sched_fair.c index 8e1352c75557..bdf64346b4d1 100644 --- a/kernel/sched_fair.c +++ b/kernel/sched_fair.c @@ -1295,16 +1295,63 @@ out: } #endif /* CONFIG_SMP */ -static unsigned long wakeup_gran(struct sched_entity *se) +/* + * 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); + /* - * More easily preempt - nice tasks, while not making it harder for - * + nice tasks. + * Since its curr running now, convert the gran from real-time + * to virtual-time in his units. */ - if (!sched_feat(ASYM_GRAN) || se->load.weight > NICE_0_LOAD) - gran = calc_delta_fair(sysctl_sched_wakeup_granularity, se); + 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); + } return gran; } @@ -1331,7 +1378,7 @@ wakeup_preempt_entity(struct sched_entity *curr, struct sched_entity *se) if (vdiff <= 0) return -1; - gran = wakeup_gran(curr); + gran = wakeup_gran(curr, se); if (vdiff > gran) return 1; diff --git a/kernel/sched_features.h b/kernel/sched_features.h index da5d93b5d2c6..76f61756e677 100644 --- a/kernel/sched_features.h +++ b/kernel/sched_features.h @@ -1,5 +1,6 @@ SCHED_FEAT(NEW_FAIR_SLEEPERS, 1) -SCHED_FEAT(NORMALIZED_SLEEPER, 1) +SCHED_FEAT(NORMALIZED_SLEEPER, 0) +SCHED_FEAT(ADAPTIVE_GRAN, 1) SCHED_FEAT(WAKEUP_PREEMPT, 1) SCHED_FEAT(START_DEBIT, 1) SCHED_FEAT(AFFINE_WAKEUPS, 1) From 34cb61359b503d7aff6447acb037a5efd6ce93b2 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Fri, 16 Jan 2009 13:36:06 +0100 Subject: [PATCH 16/46] sched: fix !CONFIG_SCHEDSTATS build failure Stephen Rothwell reported this linux-next build failure with !CONFIG_SCHEDSTATS: | In file included from kernel/sched.c:1703: | kernel/sched_fair.c: In function 'adaptive_gran': | kernel/sched_fair.c:1324: error: 'struct sched_entity' has no member named 'avg_wakeup' The start_runtime and avg_wakeup metrics are now not just for statistics, but also for scheduling - so they always need to be available. (Also move out the nr_migrations fields - for future perfcounters usage.) Reported-by: Stephen Rothwell Signed-off-by: Ingo Molnar --- include/linux/sched.h | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/include/linux/sched.h b/include/linux/sched.h index daf4e07bc978..5d56b54350a5 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -1031,6 +1031,10 @@ struct sched_entity { u64 last_wakeup; u64 avg_overlap; + u64 start_runtime; + u64 avg_wakeup; + u64 nr_migrations; + #ifdef CONFIG_SCHEDSTATS u64 wait_start; u64 wait_max; @@ -1046,10 +1050,6 @@ struct sched_entity { u64 exec_max; u64 slice_max; - u64 start_runtime; - u64 avg_wakeup; - - u64 nr_migrations; u64 nr_migrations_cold; u64 nr_failed_migrations_affine; u64 nr_failed_migrations_running; From ceacc2c1c85ac498ca4cf297bdfe5b4aaa9fd0e0 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Fri, 16 Jan 2009 14:46:40 +0100 Subject: [PATCH 17/46] sched: make plist a library facility Ingo Molnar wrote: > here's a new build failure with tip/sched/rt: > > LD .tmp_vmlinux1 > kernel/built-in.o: In function `set_curr_task_rt': > sched.c:(.text+0x3675): undefined reference to `plist_del' > kernel/built-in.o: In function `pick_next_task_rt': > sched.c:(.text+0x37ce): undefined reference to `plist_del' > kernel/built-in.o: In function `enqueue_pushable_task': > sched.c:(.text+0x381c): undefined reference to `plist_del' Eliminate the plist library kconfig and make it available unconditionally. Signed-off-by: Peter Zijlstra Signed-off-by: Ingo Molnar --- init/Kconfig | 1 - kernel/sched_rt.c | 21 +++++++++++++++------ lib/Kconfig | 6 ------ lib/Makefile | 4 ++-- 4 files changed, 17 insertions(+), 15 deletions(-) diff --git a/init/Kconfig b/init/Kconfig index a724a149bf3f..19b78aa010e3 100644 --- a/init/Kconfig +++ b/init/Kconfig @@ -874,7 +874,6 @@ config SLABINFO config RT_MUTEXES boolean - select PLIST config BASE_SMALL int diff --git a/kernel/sched_rt.c b/kernel/sched_rt.c index 4230b15fe90e..48d1f6e8497a 100644 --- a/kernel/sched_rt.c +++ b/kernel/sched_rt.c @@ -114,14 +114,23 @@ static void dequeue_pushable_task(struct rq *rq, struct task_struct *p) #else +static inline void enqueue_pushable_task(struct rq *rq, struct task_struct *p) +{ +} + +static inline void dequeue_pushable_task(struct rq *rq, struct task_struct *p) +{ +} + static inline -void enqueue_pushable_task(struct rq *rq, struct task_struct *p) {} +void inc_rt_migration(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) +{ +} + static inline -void dequeue_pushable_task(struct rq *rq, struct task_struct *p) {} -static inline -void inc_rt_migration(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) {} -static inline -void dec_rt_migration(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) {} +void dec_rt_migration(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq) +{ +} #endif /* CONFIG_SMP */ diff --git a/lib/Kconfig b/lib/Kconfig index 03c2c24b9083..fc8ea1ca59d8 100644 --- a/lib/Kconfig +++ b/lib/Kconfig @@ -136,12 +136,6 @@ config TEXTSEARCH_BM config TEXTSEARCH_FSM tristate -# -# plist support is select#ed if needed -# -config PLIST - boolean - config HAS_IOMEM boolean depends on !NO_IOMEM diff --git a/lib/Makefile b/lib/Makefile index 32b0e64ded27..902d73851044 100644 --- a/lib/Makefile +++ b/lib/Makefile @@ -11,7 +11,8 @@ lib-y := ctype.o string.o vsprintf.o cmdline.o \ rbtree.o radix-tree.o dump_stack.o \ idr.o int_sqrt.o extable.o prio_tree.o \ sha1.o irq_regs.o reciprocal_div.o argv_split.o \ - proportions.o prio_heap.o ratelimit.o show_mem.o is_single_threaded.o + proportions.o prio_heap.o ratelimit.o show_mem.o \ + is_single_threaded.o plist.o lib-$(CONFIG_MMU) += ioremap.o lib-$(CONFIG_SMP) += cpumask.o @@ -40,7 +41,6 @@ lib-$(CONFIG_GENERIC_FIND_NEXT_BIT) += find_next_bit.o lib-$(CONFIG_GENERIC_FIND_LAST_BIT) += find_last_bit.o obj-$(CONFIG_GENERIC_HWEIGHT) += hweight.o obj-$(CONFIG_LOCK_KERNEL) += kernel_lock.o -obj-$(CONFIG_PLIST) += plist.o obj-$(CONFIG_DEBUG_PREEMPT) += smp_processor_id.o obj-$(CONFIG_DEBUG_LIST) += list_debug.o obj-$(CONFIG_DEBUG_OBJECTS) += debugobjects.o From ad0b0fd554dfc126b5750d14908dccc3bbf602be Mon Sep 17 00:00:00 2001 From: Arjan van de Ven Date: Tue, 10 Feb 2009 11:42:26 -0800 Subject: [PATCH 18/46] sched, latencytop: incorporate review feedback from Andrew Morton Andrew had some suggestions for the latencytop file; this patch takes care of most of these: * Add documentation * Turn account_scheduler_latency into an inline function * Don't report negative values to userspace * Make the file operations struct const * Fix a few checkpatch.pl warnings Signed-off-by: Arjan van de Ven Signed-off-by: Andrew Morton Signed-off-by: Ingo Molnar --- include/linux/latencytop.h | 10 ++++- kernel/latencytop.c | 83 ++++++++++++++++++++++++++++++++------ 2 files changed, 80 insertions(+), 13 deletions(-) diff --git a/include/linux/latencytop.h b/include/linux/latencytop.h index 901c2d6377a8..b0e99898527c 100644 --- a/include/linux/latencytop.h +++ b/include/linux/latencytop.h @@ -9,6 +9,7 @@ #ifndef _INCLUDE_GUARD_LATENCYTOP_H_ #define _INCLUDE_GUARD_LATENCYTOP_H_ +#include #ifdef CONFIG_LATENCYTOP #define LT_SAVECOUNT 32 @@ -24,7 +25,14 @@ struct latency_record { struct task_struct; -void account_scheduler_latency(struct task_struct *task, int usecs, int inter); +extern int latencytop_enabled; +void __account_scheduler_latency(struct task_struct *task, int usecs, int inter); +static inline void +account_scheduler_latency(struct task_struct *task, int usecs, int inter) +{ + if (unlikely(latencytop_enabled)) + __account_scheduler_latency(task, usecs, inter); +} void clear_all_latency_tracing(struct task_struct *p); diff --git a/kernel/latencytop.c b/kernel/latencytop.c index 449db466bdbc..ca07c5c0c914 100644 --- a/kernel/latencytop.c +++ b/kernel/latencytop.c @@ -9,6 +9,44 @@ * as published by the Free Software Foundation; version 2 * of the License. */ + +/* + * CONFIG_LATENCYTOP enables a kernel latency tracking infrastructure that is + * used by the "latencytop" userspace tool. The latency that is tracked is not + * the 'traditional' interrupt latency (which is primarily caused by something + * else consuming CPU), but instead, it is the latency an application encounters + * because the kernel sleeps on its behalf for various reasons. + * + * This code tracks 2 levels of statistics: + * 1) System level latency + * 2) Per process latency + * + * The latency is stored in fixed sized data structures in an accumulated form; + * if the "same" latency cause is hit twice, this will be tracked as one entry + * in the data structure. Both the count, total accumulated latency and maximum + * latency are tracked in this data structure. When the fixed size structure is + * full, no new causes are tracked until the buffer is flushed by writing to + * the /proc file; the userspace tool does this on a regular basis. + * + * A latency cause is identified by a stringified backtrace at the point that + * the scheduler gets invoked. The userland tool will use this string to + * identify the cause of the latency in human readable form. + * + * The information is exported via /proc/latency_stats and /proc//latency. + * These files look like this: + * + * Latency Top version : v0.1 + * 70 59433 4897 i915_irq_wait drm_ioctl vfs_ioctl do_vfs_ioctl sys_ioctl + * | | | | + * | | | +----> the stringified backtrace + * | | +---------> The maximum latency for this entry in microseconds + * | +--------------> The accumulated latency for this entry (microseconds) + * +-------------------> The number of times this entry is hit + * + * (note: the average latency is the accumulated latency divided by the number + * of times) + */ + #include #include #include @@ -72,7 +110,7 @@ account_global_scheduler_latency(struct task_struct *tsk, struct latency_record firstnonnull = i; continue; } - for (q = 0 ; q < LT_BACKTRACEDEPTH ; q++) { + for (q = 0; q < LT_BACKTRACEDEPTH; q++) { unsigned long record = lat->backtrace[q]; if (latency_record[i].backtrace[q] != record) { @@ -101,31 +139,52 @@ account_global_scheduler_latency(struct task_struct *tsk, struct latency_record memcpy(&latency_record[i], lat, sizeof(struct latency_record)); } -static inline void store_stacktrace(struct task_struct *tsk, struct latency_record *lat) +/* + * Iterator to store a backtrace into a latency record entry + */ +static inline void store_stacktrace(struct task_struct *tsk, + struct latency_record *lat) { struct stack_trace trace; memset(&trace, 0, sizeof(trace)); trace.max_entries = LT_BACKTRACEDEPTH; trace.entries = &lat->backtrace[0]; - trace.skip = 0; save_stack_trace_tsk(tsk, &trace); } +/** + * __account_scheduler_latency - record an occured latency + * @tsk - the task struct of the task hitting the latency + * @usecs - the duration of the latency in microseconds + * @inter - 1 if the sleep was interruptible, 0 if uninterruptible + * + * This function is the main entry point for recording latency entries + * as called by the scheduler. + * + * This function has a few special cases to deal with normal 'non-latency' + * sleeps: specifically, interruptible sleep longer than 5 msec is skipped + * since this usually is caused by waiting for events via select() and co. + * + * Negative latencies (caused by time going backwards) are also explicitly + * skipped. + */ void __sched -account_scheduler_latency(struct task_struct *tsk, int usecs, int inter) +__account_scheduler_latency(struct task_struct *tsk, int usecs, int inter) { unsigned long flags; int i, q; struct latency_record lat; - if (!latencytop_enabled) - return; - /* Long interruptible waits are generally user requested... */ if (inter && usecs > 5000) return; + /* Negative sleeps are time going backwards */ + /* Zero-time sleeps are non-interesting */ + if (usecs <= 0) + return; + memset(&lat, 0, sizeof(lat)); lat.count = 1; lat.time = usecs; @@ -143,12 +202,12 @@ account_scheduler_latency(struct task_struct *tsk, int usecs, int inter) if (tsk->latency_record_count >= LT_SAVECOUNT) goto out_unlock; - for (i = 0; i < LT_SAVECOUNT ; i++) { + for (i = 0; i < LT_SAVECOUNT; i++) { struct latency_record *mylat; int same = 1; mylat = &tsk->latency_record[i]; - for (q = 0 ; q < LT_BACKTRACEDEPTH ; q++) { + for (q = 0; q < LT_BACKTRACEDEPTH; q++) { unsigned long record = lat.backtrace[q]; if (mylat->backtrace[q] != record) { @@ -186,7 +245,7 @@ static int lstats_show(struct seq_file *m, void *v) for (i = 0; i < MAXLR; i++) { if (latency_record[i].backtrace[0]) { int q; - seq_printf(m, "%i %li %li ", + seq_printf(m, "%i %lu %lu ", latency_record[i].count, latency_record[i].time, latency_record[i].max); @@ -223,7 +282,7 @@ static int lstats_open(struct inode *inode, struct file *filp) return single_open(filp, lstats_show, NULL); } -static struct file_operations lstats_fops = { +static const struct file_operations lstats_fops = { .open = lstats_open, .read = seq_read, .write = lstats_write, @@ -236,4 +295,4 @@ static int __init init_lstats_procfs(void) proc_create("latency_stats", 0644, NULL, &lstats_fops); return 0; } -__initcall(init_lstats_procfs); +device_initcall(init_lstats_procfs); From a0a522ce3d6d8c907e45d4f2730ee8573484cc88 Mon Sep 17 00:00:00 2001 From: Henrik Austad Date: Fri, 13 Feb 2009 20:35:45 +0100 Subject: [PATCH 19/46] sched: idle_at_tick is only used when CONFIG_SMP is set Impact: struct rq size optimization The idle_at_tick in struct rq is only used in SMP settings and it does not make sense to have this in the rq in an UP setup. Signed-off-by: Henrik Austad Signed-off-by: Ingo Molnar --- kernel/sched.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/kernel/sched.c b/kernel/sched.c index 5faf5d482fcd..648154cf1117 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -555,7 +555,6 @@ struct rq { unsigned long nr_running; #define CPU_LOAD_IDX_MAX 5 unsigned long cpu_load[CPU_LOAD_IDX_MAX]; - unsigned char idle_at_tick; #ifdef CONFIG_NO_HZ unsigned long last_tick_seen; unsigned char in_nohz_recently; @@ -596,6 +595,7 @@ struct rq { struct root_domain *rd; struct sched_domain *sd; + unsigned char idle_at_tick; /* For active balancing */ int active_balance; int push_cpu; From 2b8f836fb196acede88b6cc772e9057e0a9c0223 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Am=C3=A9rico=20Wang?= Date: Mon, 16 Feb 2009 18:54:21 +0800 Subject: [PATCH 20/46] sched: use TASK_NICE for task_struct #define TASK_NICE(p) PRIO_TO_NICE((p)->static_prio) So it's better to use TASK_NICE here. Signed-off-by: WANG Cong Acked-by: Peter Zijlstra Signed-off-by: Ingo Molnar --- kernel/sched.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/kernel/sched.c b/kernel/sched.c index 648154cf1117..5475d56a20f1 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -5236,7 +5236,7 @@ SYSCALL_DEFINE1(nice, int, increment) if (increment > 40) increment = 40; - nice = PRIO_TO_NICE(current->static_prio) + increment; + nice = TASK_NICE(current) + increment; if (nice < -20) nice = -20; if (nice > 19) From c40c6f85a7594ad842233885386a0ca4cd40eafe Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Thu, 26 Feb 2009 15:40:15 +0800 Subject: [PATCH 21/46] cpuacct: add a branch prediction cpuacct_charge() is in fast-path, and checking of !cpuacct_susys.active always returns false after cpuacct has been initialized at system boot. Signed-off-by: Li Zefan Cc: Peter Zijlstra Cc: Paul Menage Cc: Balbir Singh Signed-off-by: Ingo Molnar --- kernel/sched.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/kernel/sched.c b/kernel/sched.c index 5475d56a20f1..8e63ffb6ed05 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -9684,7 +9684,7 @@ static void cpuacct_charge(struct task_struct *tsk, u64 cputime) struct cpuacct *ca; int cpu; - if (!cpuacct_subsys.active) + if (unlikely(!cpuacct_subsys.active)) return; cpu = task_cpu(tsk); From b342501cd31e5546d0c9ca8ceff5ded1832f9e5b Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Thu, 26 Feb 2009 20:20:29 +0100 Subject: [PATCH 22/46] sched: allow architectures to specify sched_clock_stable Allow CONFIG_HAVE_UNSTABLE_SCHED_CLOCK architectures to still specify that their sched_clock() implementation is reliable. This will be used by x86 to switch on a faster sched_clock_cpu() implementation on certain CPU types. Signed-off-by: Ingo Molnar --- include/linux/sched.h | 10 ++++++++++ kernel/sched_clock.c | 45 +++++++++++++++++++------------------------ 2 files changed, 30 insertions(+), 25 deletions(-) diff --git a/include/linux/sched.h b/include/linux/sched.h index 8981e52c714f..a063d19b7a7d 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -1670,6 +1670,16 @@ static inline int set_cpus_allowed(struct task_struct *p, cpumask_t new_mask) return set_cpus_allowed_ptr(p, &new_mask); } +/* + * Architectures can set this to 1 if they have specified + * CONFIG_HAVE_UNSTABLE_SCHED_CLOCK in their arch Kconfig, + * but then during bootup it turns out that sched_clock() + * is reliable after all: + */ +#ifdef CONFIG_HAVE_UNSTABLE_SCHED_CLOCK +extern int sched_clock_stable; +#endif + extern unsigned long long sched_clock(void); extern void sched_clock_init(void); diff --git a/kernel/sched_clock.c b/kernel/sched_clock.c index a0b0852414cc..a755d023805a 100644 --- a/kernel/sched_clock.c +++ b/kernel/sched_clock.c @@ -24,11 +24,11 @@ * The clock: sched_clock_cpu() is monotonic per cpu, and should be somewhat * consistent between cpus (never more than 2 jiffies difference). */ -#include -#include #include -#include #include +#include +#include +#include /* * Scheduler clock - returns current time in nanosec units. @@ -43,6 +43,10 @@ unsigned long long __attribute__((weak)) sched_clock(void) static __read_mostly int sched_clock_running; #ifdef CONFIG_HAVE_UNSTABLE_SCHED_CLOCK +__read_mostly int sched_clock_stable; +#else +static const int sched_clock_stable = 1; +#endif struct sched_clock_data { /* @@ -87,7 +91,7 @@ void sched_clock_init(void) } /* - * min,max except they take wrapping into account + * min, max except they take wrapping into account */ static inline u64 wrap_min(u64 x, u64 y) @@ -116,10 +120,13 @@ static u64 __update_sched_clock(struct sched_clock_data *scd, u64 now) if (unlikely(delta < 0)) delta = 0; + if (unlikely(!sched_clock_running)) + return 0ull; + /* * scd->clock = clamp(scd->tick_gtod + delta, - * max(scd->tick_gtod, scd->clock), - * scd->tick_gtod + TICK_NSEC); + * max(scd->tick_gtod, scd->clock), + * scd->tick_gtod + TICK_NSEC); */ clock = scd->tick_gtod + delta; @@ -148,12 +155,13 @@ static void lock_double_clock(struct sched_clock_data *data1, u64 sched_clock_cpu(int cpu) { - struct sched_clock_data *scd = cpu_sdc(cpu); u64 now, clock, this_clock, remote_clock; + struct sched_clock_data *scd; - if (unlikely(!sched_clock_running)) - return 0ull; + if (sched_clock_stable) + return sched_clock(); + scd = cpu_sdc(cpu); WARN_ON_ONCE(!irqs_disabled()); now = sched_clock(); @@ -193,6 +201,8 @@ u64 sched_clock_cpu(int cpu) return clock; } +#ifdef CONFIG_HAVE_UNSTABLE_SCHED_CLOCK + void sched_clock_tick(void) { struct sched_clock_data *scd = this_scd(); @@ -235,22 +245,7 @@ void sched_clock_idle_wakeup_event(u64 delta_ns) } EXPORT_SYMBOL_GPL(sched_clock_idle_wakeup_event); -#else /* CONFIG_HAVE_UNSTABLE_SCHED_CLOCK */ - -void sched_clock_init(void) -{ - sched_clock_running = 1; -} - -u64 sched_clock_cpu(int cpu) -{ - if (unlikely(!sched_clock_running)) - return 0; - - return sched_clock(); -} - -#endif +#endif /* CONFIG_HAVE_UNSTABLE_SCHED_CLOCK */ unsigned long long cpu_clock(int cpu) { From 83ce400928680a6c8123d492684b27857f5a2d95 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Thu, 26 Feb 2009 20:16:58 +0100 Subject: [PATCH 23/46] x86: set X86_FEATURE_TSC_RELIABLE If the TSC is constant and non-stop, also set it reliable. (We will turn this off in DMI quirks for multi-chassis systems) The performance number on a 16-way Nehalem system running 32 tasks that context-switch between each other is significant: sched_clock_stable=0 sched_clock_stable=1 .................... .................... 22.456925 million/sec 24.306972 million/sec [+8.2%] lmbench's "lat_ctx -s 0 2" goes from 0.63 microseconds to 0.59 microseconds - a 6.7% increase in context-switching performance. Perfstat of 1 million pipe context switches between two tasks: Performance counter stats for './pipe-test-1m': [before] [after] ............ ............ 37621.421089 36436.848378 task clock ticks (msecs) 0 0 CPU migrations (events) 2000274 2000189 context switches (events) 194 193 pagefaults (events) 8433799643 8171016416 CPU cycles (events) -3.21% 8370133368 8180999694 instructions (events) -2.31% 4158565 3895941 cache references (events) -6.74% 44312 46264 cache misses (events) 2349.287976 2279.362465 wall-time (msecs) -3.06% The speedup comes straight from the reduction in the instruction count. sched_clock_cpu() got simpler and the whole workload thus executes faster. Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/intel.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) diff --git a/arch/x86/kernel/cpu/intel.c b/arch/x86/kernel/cpu/intel.c index 24ff26a38ade..5fff00c70de0 100644 --- a/arch/x86/kernel/cpu/intel.c +++ b/arch/x86/kernel/cpu/intel.c @@ -4,6 +4,7 @@ #include #include #include +#include #include #include @@ -56,11 +57,16 @@ static void __cpuinit early_init_intel(struct cpuinfo_x86 *c) /* * c->x86_power is 8000_0007 edx. Bit 8 is TSC runs at constant rate - * with P/T states and does not stop in deep C-states + * with P/T states and does not stop in deep C-states. + * + * It is also reliable across cores and sockets. (but not across + * cabinets - we turn it off in that case explicitly.) */ if (c->x86_power & (1 << 8)) { set_cpu_cap(c, X86_FEATURE_CONSTANT_TSC); set_cpu_cap(c, X86_FEATURE_NONSTOP_TSC); + set_cpu_cap(c, X86_FEATURE_TSC_RELIABLE); + sched_clock_stable = 1; } } From 8325d9c09dedf45476f4d6261d1b6a72e4a7453f Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Thu, 26 Feb 2009 21:40:16 +0100 Subject: [PATCH 24/46] sched_clock: cleanups - remove superfluous checks in __update_sched_clock() - skip sched_clock_tick() for sched_clock_stable - reinstate the simple !HAVE_UNSTABLE_SCHED_CLOCK code to please the bloatwatch Signed-off-by: Peter Zijlstra Signed-off-by: Ingo Molnar --- kernel/sched_clock.c | 31 ++++++++++++++++++++----------- 1 file changed, 20 insertions(+), 11 deletions(-) diff --git a/kernel/sched_clock.c b/kernel/sched_clock.c index a755d023805a..390f33234bd0 100644 --- a/kernel/sched_clock.c +++ b/kernel/sched_clock.c @@ -44,9 +44,6 @@ static __read_mostly int sched_clock_running; #ifdef CONFIG_HAVE_UNSTABLE_SCHED_CLOCK __read_mostly int sched_clock_stable; -#else -static const int sched_clock_stable = 1; -#endif struct sched_clock_data { /* @@ -115,14 +112,9 @@ static u64 __update_sched_clock(struct sched_clock_data *scd, u64 now) s64 delta = now - scd->tick_raw; u64 clock, min_clock, max_clock; - WARN_ON_ONCE(!irqs_disabled()); - if (unlikely(delta < 0)) delta = 0; - if (unlikely(!sched_clock_running)) - return 0ull; - /* * scd->clock = clamp(scd->tick_gtod + delta, * max(scd->tick_gtod, scd->clock), @@ -201,18 +193,20 @@ u64 sched_clock_cpu(int cpu) return clock; } -#ifdef CONFIG_HAVE_UNSTABLE_SCHED_CLOCK - void sched_clock_tick(void) { - struct sched_clock_data *scd = this_scd(); + struct sched_clock_data *scd; u64 now, now_gtod; + if (sched_clock_stable) + return; + if (unlikely(!sched_clock_running)) return; WARN_ON_ONCE(!irqs_disabled()); + scd = this_scd(); now_gtod = ktime_to_ns(ktime_get()); now = sched_clock(); @@ -245,6 +239,21 @@ void sched_clock_idle_wakeup_event(u64 delta_ns) } EXPORT_SYMBOL_GPL(sched_clock_idle_wakeup_event); +#else /* CONFIG_HAVE_UNSTABLE_SCHED_CLOCK */ + +void sched_clock_init(void) +{ + sched_clock_running = 1; +} + +u64 sched_clock_cpu(int cpu) +{ + if (unlikely(!sched_clock_running)) + return 0; + + return sched_clock(); +} + #endif /* CONFIG_HAVE_UNSTABLE_SCHED_CLOCK */ unsigned long long cpu_clock(int cpu) From 72fd455ba54b5a02b9c74221b9ded8b1845b464a Mon Sep 17 00:00:00 2001 From: Wang Chen Date: Mon, 2 Mar 2009 13:55:14 +0800 Subject: [PATCH 25/46] sched, documentation: remove old O(1) scheduler document Since we don't have O(1) scheduler implementation anymore, remove the legacy doc. Signed-off-by: Wang Chen Signed-off-by: Ingo Molnar --- Documentation/scheduler/00-INDEX | 2 - Documentation/scheduler/sched-coding.txt | 126 ----------------------- 2 files changed, 128 deletions(-) delete mode 100644 Documentation/scheduler/sched-coding.txt diff --git a/Documentation/scheduler/00-INDEX b/Documentation/scheduler/00-INDEX index aabcc3a089ba..3c00c9c3219e 100644 --- a/Documentation/scheduler/00-INDEX +++ b/Documentation/scheduler/00-INDEX @@ -2,8 +2,6 @@ - this file. sched-arch.txt - CPU Scheduler implementation hints for architecture specific code. -sched-coding.txt - - reference for various scheduler-related methods in the O(1) scheduler. sched-design-CFS.txt - goals, design and implementation of the Complete Fair Scheduler. sched-domains.txt diff --git a/Documentation/scheduler/sched-coding.txt b/Documentation/scheduler/sched-coding.txt deleted file mode 100644 index cbd8db752acf..000000000000 --- a/Documentation/scheduler/sched-coding.txt +++ /dev/null @@ -1,126 +0,0 @@ - Reference for various scheduler-related methods in the O(1) scheduler - Robert Love , MontaVista Software - - -Note most of these methods are local to kernel/sched.c - this is by design. -The scheduler is meant to be self-contained and abstracted away. This document -is primarily for understanding the scheduler, not interfacing to it. Some of -the discussed interfaces, however, are general process/scheduling methods. -They are typically defined in include/linux/sched.h. - - -Main Scheduling Methods ------------------------ - -void load_balance(runqueue_t *this_rq, int idle) - Attempts to pull tasks from one cpu to another to balance cpu usage, - if needed. This method is called explicitly if the runqueues are - imbalanced or periodically by the timer tick. Prior to calling, - the current runqueue must be locked and interrupts disabled. - -void schedule() - The main scheduling function. Upon return, the highest priority - process will be active. - - -Locking -------- - -Each runqueue has its own lock, rq->lock. When multiple runqueues need -to be locked, lock acquires must be ordered by ascending &runqueue value. - -A specific runqueue is locked via - - task_rq_lock(task_t pid, unsigned long *flags) - -which disables preemption, disables interrupts, and locks the runqueue pid is -running on. Likewise, - - task_rq_unlock(task_t pid, unsigned long *flags) - -unlocks the runqueue pid is running on, restores interrupts to their previous -state, and reenables preemption. - -The routines - - double_rq_lock(runqueue_t *rq1, runqueue_t *rq2) - -and - - double_rq_unlock(runqueue_t *rq1, runqueue_t *rq2) - -safely lock and unlock, respectively, the two specified runqueues. They do -not, however, disable and restore interrupts. Users are required to do so -manually before and after calls. - - -Values ------- - -MAX_PRIO - The maximum priority of the system, stored in the task as task->prio. - Lower priorities are higher. Normal (non-RT) priorities range from - MAX_RT_PRIO to (MAX_PRIO - 1). -MAX_RT_PRIO - The maximum real-time priority of the system. Valid RT priorities - range from 0 to (MAX_RT_PRIO - 1). -MAX_USER_RT_PRIO - The maximum real-time priority that is exported to user-space. Should - always be equal to or less than MAX_RT_PRIO. Setting it less allows - kernel threads to have higher priorities than any user-space task. -MIN_TIMESLICE -MAX_TIMESLICE - Respectively, the minimum and maximum timeslices (quanta) of a process. - -Data ----- - -struct runqueue - The main per-CPU runqueue data structure. -struct task_struct - The main per-process data structure. - - -General Methods ---------------- - -cpu_rq(cpu) - Returns the runqueue of the specified cpu. -this_rq() - Returns the runqueue of the current cpu. -task_rq(pid) - Returns the runqueue which holds the specified pid. -cpu_curr(cpu) - Returns the task currently running on the given cpu. -rt_task(pid) - Returns true if pid is real-time, false if not. - - -Process Control Methods ------------------------ - -void set_user_nice(task_t *p, long nice) - Sets the "nice" value of task p to the given value. -int setscheduler(pid_t pid, int policy, struct sched_param *param) - Sets the scheduling policy and parameters for the given pid. -int set_cpus_allowed(task_t *p, unsigned long new_mask) - Sets a given task's CPU affinity and migrates it to a proper cpu. - Callers must have a valid reference to the task and assure the - task not exit prematurely. No locks can be held during the call. -set_task_state(tsk, state_value) - Sets the given task's state to the given value. -set_current_state(state_value) - Sets the current task's state to the given value. -void set_tsk_need_resched(struct task_struct *tsk) - Sets need_resched in the given task. -void clear_tsk_need_resched(struct task_struct *tsk) - Clears need_resched in the given task. -void set_need_resched() - Sets need_resched in the current task. -void clear_need_resched() - Clears need_resched in the current task. -int need_resched() - Returns true if need_resched is set in the current task, false - otherwise. -yield() - Place the current process at the end of the runqueue and call schedule. From b67802ea8061393f7bd2d4db934646e76096027c Mon Sep 17 00:00:00 2001 From: Wang Chen Date: Mon, 2 Mar 2009 13:55:26 +0800 Subject: [PATCH 26/46] sched: kill unused parameter of pick_next_task() Impact: micro-optimization Parameter "prev" is not used really. Signed-off-by: Wang Chen Signed-off-by: Ingo Molnar --- kernel/sched.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/kernel/sched.c b/kernel/sched.c index dfae1bf6d5b2..9fe8e17574af 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -4603,7 +4603,7 @@ 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) { const struct sched_class *class; struct task_struct *p; @@ -4678,7 +4678,7 @@ need_resched_nonpreemptible: idle_balance(cpu, rq); prev->sched_class->put_prev_task(rq, prev); - next = pick_next_task(rq, prev); + next = pick_next_task(rq); if (likely(prev != next)) { sched_info_switch(prev, next); @@ -6514,7 +6514,7 @@ static void migrate_dead_tasks(unsigned int dead_cpu) if (!rq->nr_running) break; update_rq_clock(rq); - next = pick_next_task(rq, rq->curr); + next = pick_next_task(rq); if (!next) break; next->sched_class->put_prev_task(rq, next); From 8a0be9ef8225638d26b455788f988c8f84ce9e75 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Thu, 5 Mar 2009 01:27:02 +0100 Subject: [PATCH 27/46] sched: don't rebalance if attached on NULL domain Impact: fix function graph trace hang / drop pointless softirq on UP While debugging a function graph trace hang on an old PII, I saw that it consumed most of its time on the timer interrupt. And the domain rebalancing softirq was the most concerned. The timer interrupt calls trigger_load_balance() which will decide if it is worth to schedule a rebalancing softirq. In case of builtin UP kernel, no problem arises because there is no domain question. In case of builtin SMP kernel running on an SMP box, still no problem, the softirq will be raised each time we reach the next_balance time. In case of builtin SMP kernel running on a UP box (most distros provide default SMP kernels, whatever the box you have), then the CPU is attached to the NULL sched domain. So a kind of unexpected behaviour happen: trigger_load_balance() -> raises the rebalancing softirq later on softirq: run_rebalance_domains() -> rebalance_domains() where the for_each_domain(cpu, sd) is not taken because of the NULL domain we are attached at. Which means rq->next_balance is never updated. So on the next timer tick, we will enter trigger_load_balance() which will always reschedule() the rebalacing softirq: if (time_after_eq(jiffies, rq->next_balance)) raise_softirq(SCHED_SOFTIRQ); So for each tick, we process this pointless softirq. This patch fixes it by checking if we are attached to the null domain before raising the softirq, another possible fix would be to set the maximal possible JIFFIES value to rq->next_balance if we are attached to the NULL domain. v2: build fix on UP Signed-off-by: Frederic Weisbecker Cc: Steven Rostedt Cc: Peter Zijlstra LKML-Reference: <49af242d.1c07d00a.32d5.ffffc019@mx.google.com> Signed-off-by: Ingo Molnar --- kernel/sched.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) diff --git a/kernel/sched.c b/kernel/sched.c index dfae1bf6d5b2..e509dbd7d77f 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -4148,6 +4148,11 @@ static void run_rebalance_domains(struct softirq_action *h) #endif } +static inline int on_null_domain(int cpu) +{ + return !rcu_dereference(cpu_rq(cpu)->sd); +} + /* * Trigger the SCHED_SOFTIRQ if it is time to do periodic load balancing. * @@ -4205,7 +4210,9 @@ static inline void trigger_load_balance(struct rq *rq, int cpu) cpumask_test_cpu(cpu, nohz.cpu_mask)) return; #endif - if (time_after_eq(jiffies, rq->next_balance)) + /* 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); } From 5ed0cec0ac5f1b3759bdbe4d9df32ee4ff8afb5a Mon Sep 17 00:00:00 2001 From: Lai Jiangshan Date: Fri, 6 Mar 2009 19:40:20 +0800 Subject: [PATCH 28/46] sched: TIF_NEED_RESCHED -> need_reshed() cleanup Impact: cleanup Use test_tsk_need_resched(), set_tsk_need_resched(), need_resched() instead of using TIF_NEED_RESCHED. Signed-off-by: Lai Jiangshan Cc: Peter Zijlstra LKML-Reference: <49B10BA4.9070209@cn.fujitsu.com> Signed-off-by: Ingo Molnar --- kernel/sched.c | 10 +++++----- lib/kernel_lock.c | 2 +- 2 files changed, 6 insertions(+), 6 deletions(-) diff --git a/kernel/sched.c b/kernel/sched.c index 8b92f40c147d..e0fa739a441b 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -1189,10 +1189,10 @@ static void resched_task(struct task_struct *p) assert_spin_locked(&task_rq(p)->lock); - if (unlikely(test_tsk_thread_flag(p, TIF_NEED_RESCHED))) + if (test_tsk_need_resched(p)) return; - set_tsk_thread_flag(p, TIF_NEED_RESCHED); + set_tsk_need_resched(p); cpu = task_cpu(p); if (cpu == smp_processor_id()) @@ -1248,7 +1248,7 @@ void wake_up_idle_cpu(int cpu) * lockless. The worst case is that the other CPU runs the * idle task through an additional NOOP schedule() */ - set_tsk_thread_flag(rq->idle, TIF_NEED_RESCHED); + set_tsk_need_resched(rq->idle); /* NEED_RESCHED must be visible before we test polling */ smp_mb(); @@ -4740,7 +4740,7 @@ asmlinkage void __sched preempt_schedule(void) * between schedule and now. */ barrier(); - } while (unlikely(test_thread_flag(TIF_NEED_RESCHED))); + } while (need_resched()); } EXPORT_SYMBOL(preempt_schedule); @@ -4769,7 +4769,7 @@ asmlinkage void __sched preempt_schedule_irq(void) * between schedule and now. */ barrier(); - } while (unlikely(test_thread_flag(TIF_NEED_RESCHED))); + } while (need_resched()); } #endif /* CONFIG_PREEMPT */ diff --git a/lib/kernel_lock.c b/lib/kernel_lock.c index 01a3c22c1b5a..39f1029e3525 100644 --- a/lib/kernel_lock.c +++ b/lib/kernel_lock.c @@ -39,7 +39,7 @@ static __cacheline_aligned_in_smp DEFINE_SPINLOCK(kernel_flag); int __lockfunc __reacquire_kernel_lock(void) { while (!_raw_spin_trylock(&kernel_flag)) { - if (test_thread_flag(TIF_NEED_RESCHED)) + if (need_resched()) return -EAGAIN; cpu_relax(); } From 57310a98a354e84279d7c8af2f48805a62372e53 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Mon, 9 Mar 2009 13:56:21 +0100 Subject: [PATCH 29/46] sched: optimize ttwu vs group scheduling Impact: micro-optimization We can avoid the sched domain walk on try_to_wake_up() when we know there are no groups. Signed-off-by: Peter Zijlstra LKML-Reference: <1236603381.8389.455.camel@laptop> Signed-off-by: Ingo Molnar --- kernel/sched.c | 16 +++++++++++++++- 1 file changed, 15 insertions(+), 1 deletion(-) diff --git a/kernel/sched.c b/kernel/sched.c index e0fa739a441b..af5cd1b2d03e 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -331,6 +331,13 @@ static DEFINE_PER_CPU(struct rt_rq, init_rt_rq) ____cacheline_aligned_in_smp; */ static DEFINE_SPINLOCK(task_group_lock); +#ifdef CONFIG_SMP +static int root_task_group_empty(void) +{ + return list_empty(&root_task_group.children); +} +#endif + #ifdef CONFIG_FAIR_GROUP_SCHED #ifdef CONFIG_USER_SCHED # define INIT_TASK_GROUP_LOAD (2*NICE_0_LOAD) @@ -391,6 +398,13 @@ static inline void set_task_rq(struct task_struct *p, unsigned int cpu) #else +#ifdef CONFIG_SMP +static int root_task_group_empty(void) +{ + return 1; +} +#endif + static inline void set_task_rq(struct task_struct *p, unsigned int cpu) { } static inline struct task_group *task_group(struct task_struct *p) { @@ -2318,7 +2332,7 @@ static int try_to_wake_up(struct task_struct *p, unsigned int state, int sync) sync = 0; #ifdef CONFIG_SMP - if (sched_feat(LB_WAKEUP_UPDATE)) { + if (sched_feat(LB_WAKEUP_UPDATE) && !root_task_group_empty()) { struct sched_domain *sd; this_cpu = raw_smp_processor_id(); From f24ade3a3332811a512ed3b6c6aa69486719b1d8 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Tue, 10 Mar 2009 19:02:30 +0100 Subject: [PATCH 30/46] x86, sched_clock(): mark variables read-mostly Impact: micro-optimization There's a number of variables in the sched_clock() path that are in .data/.bss - but not marked __read_mostly. This creates the danger of accidental false cacheline sharing with some other, write-often variable. So mark them __read_mostly. Signed-off-by: Ingo Molnar --- arch/x86/kernel/tsc.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) diff --git a/arch/x86/kernel/tsc.c b/arch/x86/kernel/tsc.c index 599e58168631..9c8b71531ca8 100644 --- a/arch/x86/kernel/tsc.c +++ b/arch/x86/kernel/tsc.c @@ -17,20 +17,21 @@ #include #include -unsigned int cpu_khz; /* TSC clocks / usec, not used here */ +unsigned int __read_mostly cpu_khz; /* TSC clocks / usec, not used here */ EXPORT_SYMBOL(cpu_khz); -unsigned int tsc_khz; + +unsigned int __read_mostly tsc_khz; EXPORT_SYMBOL(tsc_khz); /* * TSC can be unstable due to cpufreq or due to unsynced TSCs */ -static int tsc_unstable; +static int __read_mostly tsc_unstable; /* native_sched_clock() is called before tsc_init(), so we must start with the TSC soft disabled to prevent erroneous rdtsc usage on !cpu_has_tsc processors */ -static int tsc_disabled = -1; +static int __read_mostly tsc_disabled = -1; static int tsc_clocksource_reliable; /* From df1c99d416500da8d26a4d78777467c53ee7689e Mon Sep 17 00:00:00 2001 From: Mike Galbraith Date: Tue, 10 Mar 2009 19:08:11 +0100 Subject: [PATCH 31/46] sched: add avg_overlap decay Impact: more precise avg_overlap metric - better load-balancing avg_overlap is used to measure the runtime overlap of the waker and wakee. However, when a process changes behaviour, eg a pipe becomes un-congested and we don't need to go to sleep after a wakeup for a while, the avg_overlap value grows stale. When running we use the avg runtime between preemption as a measure for avg_overlap since the amount of runtime can be correlated to cache footprint. The longer we run, the less likely we'll be wanting to be migrated to another CPU. Signed-off-by: Mike Galbraith Signed-off-by: Peter Zijlstra LKML-Reference: <1236709131.25234.576.camel@laptop> Signed-off-by: Ingo Molnar --- kernel/sched.c | 24 +++++++++++++++++++++++- 1 file changed, 23 insertions(+), 1 deletion(-) diff --git a/kernel/sched.c b/kernel/sched.c index af5cd1b2d03e..2f28351892c9 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -4620,6 +4620,28 @@ static inline void schedule_debug(struct task_struct *prev) #endif } +static void put_prev_task(struct rq *rq, struct task_struct *prev) +{ + if (prev->state == TASK_RUNNING) { + u64 runtime = prev->se.sum_exec_runtime; + + runtime -= prev->se.prev_sum_exec_runtime; + runtime = min_t(u64, runtime, 2*sysctl_sched_migration_cost); + + /* + * In order to avoid avg_overlap growing stale when we are + * indeed overlapping and hence not getting put to sleep, grow + * the avg_overlap on preemption. + * + * We use the average preemption runtime because that + * correlates to the amount of cache footprint a task can + * build up. + */ + update_avg(&prev->se.avg_overlap, runtime); + } + prev->sched_class->put_prev_task(rq, prev); +} + /* * Pick up the highest-prio task: */ @@ -4698,7 +4720,7 @@ need_resched_nonpreemptible: if (unlikely(!rq->nr_running)) idle_balance(cpu, rq); - prev->sched_class->put_prev_task(rq, prev); + put_prev_task(rq, prev); next = pick_next_task(rq); if (likely(prev != next)) { From 80dd99b368cf6501be88ab517bbbb5bf352b75b8 Mon Sep 17 00:00:00 2001 From: Luis Henriques Date: Mon, 16 Mar 2009 19:58:09 +0000 Subject: [PATCH 32/46] sched: fix typos in documentation Fixed typos in function documentation. Signed-off-by: Luis Henriques LKML-Reference: <20090316195809.GA6073@hades.domain.com> Signed-off-by: Ingo Molnar --- kernel/sched.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/kernel/sched.c b/kernel/sched.c index 2f28351892c9..489e7d926408 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -2082,7 +2082,7 @@ unsigned long wait_task_inactive(struct task_struct *p, long match_state) * it must be off the runqueue _entirely_, and not * preempted! * - * So if it wa still runnable (but just not actively + * So if it was still runnable (but just not actively * running right now), it's preempted, and we should * yield - it could be a while. */ @@ -2574,7 +2574,7 @@ void wake_up_new_task(struct task_struct *p, unsigned long clone_flags) #ifdef CONFIG_PREEMPT_NOTIFIERS /** - * preempt_notifier_register - tell me when current is being being preempted & rescheduled + * preempt_notifier_register - tell me when current is being preempted & rescheduled * @notifier: notifier struct to register */ void preempt_notifier_register(struct preempt_notifier *notifier) From 708dc5125309cd33c5daaad3026cc4ae6ef39c8b Mon Sep 17 00:00:00 2001 From: Luis Henriques Date: Mon, 16 Mar 2009 19:59:02 +0000 Subject: [PATCH 33/46] sched: small optimisation of can_migrate_task() There were 3 invocations of task_hot() in can_migrate_task(). Replace these 3 invocations by only one invocation, cached in a local variable. Signed-off-by: Luis Henriques LKML-Reference: <20090316195902.GA6197@hades.domain.com> Signed-off-by: Ingo Molnar --- kernel/sched.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) diff --git a/kernel/sched.c b/kernel/sched.c index 489e7d926408..d2dfe4c1a225 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -3002,6 +3002,7 @@ int can_migrate_task(struct task_struct *p, struct rq *rq, int this_cpu, struct sched_domain *sd, enum cpu_idle_type idle, int *all_pinned) { + int tsk_cache_hot = 0; /* * We do not migrate tasks that are: * 1) running (obviously), or @@ -3025,10 +3026,11 @@ int can_migrate_task(struct task_struct *p, struct rq *rq, int this_cpu, * 2) too many balance attempts have failed. */ - if (!task_hot(p, rq->clock, sd) || - sd->nr_balance_failed > sd->cache_nice_tries) { + tsk_cache_hot = task_hot(p, rq->clock, sd); + if (!tsk_cache_hot || + sd->nr_balance_failed > sd->cache_nice_tries) { #ifdef CONFIG_SCHEDSTATS - if (task_hot(p, rq->clock, sd)) { + if (tsk_cache_hot) { schedstat_inc(sd, lb_hot_gained[idle]); schedstat_inc(p, se.nr_forced_migrations); } @@ -3036,7 +3038,7 @@ int can_migrate_task(struct task_struct *p, struct rq *rq, int this_cpu, return 1; } - if (task_hot(p, rq->clock, sd)) { + if (tsk_cache_hot) { schedstat_inc(p, se.nr_failed_migrations_hot); return 0; } From af66df5ecf9c9e2d2ff86e8203510c1c4519d64c Mon Sep 17 00:00:00 2001 From: Luis Henriques Date: Wed, 18 Mar 2009 00:04:25 +0000 Subject: [PATCH 34/46] sched: jiffies not printed per CPU The jiffies value was being printed for each CPU, which does not seem to make sense. Moved jiffies to system section. Signed-off-by: Luis Henriques Acked-by: Peter Zijlstra LKML-Reference: <20090318000425.GA2228@hades.domain.com> Signed-off-by: Ingo Molnar --- kernel/sched_debug.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/kernel/sched_debug.c b/kernel/sched_debug.c index 2b1260f0e800..4daebffa0565 100644 --- a/kernel/sched_debug.c +++ b/kernel/sched_debug.c @@ -272,7 +272,6 @@ static void print_cpu(struct seq_file *m, int cpu) P(nr_switches); P(nr_load_updates); P(nr_uninterruptible); - SEQ_printf(m, " .%-30s: %lu\n", "jiffies", jiffies); PN(next_balance); P(curr->pid); PN(clock); @@ -325,6 +324,7 @@ static int sched_debug_show(struct seq_file *m, void *v) SEQ_printf(m, " .%-40s: %Ld\n", #x, (long long)(x)) #define PN(x) \ SEQ_printf(m, " .%-40s: %Ld.%06ld\n", #x, SPLIT_NS(x)) + P(jiffies); PN(sysctl_sched_latency); PN(sysctl_sched_min_granularity); PN(sysctl_sched_wakeup_granularity); From 67aa0f767af488a7f1e41cccb4f7a4893f24a1ab Mon Sep 17 00:00:00 2001 From: Luis Henriques Date: Tue, 24 Mar 2009 22:10:02 +0000 Subject: [PATCH 35/46] sched: remove unused fields from struct rq Impact: cleanup, new schedstat ABI Since they are used on in statistics and are always set to zero, the following fields from struct rq have been removed: yld_exp_empty, yld_act_empty and yld_both_empty. Both Sched Debug and SCHEDSTAT_VERSION versions has also been incremented since ABIs have been changed. The schedtop tool has been updated to properly handle new version of schedstat: http://rt.wiki.kernel.org/index.php/Schedtop_utility Signed-off-by: Luis Henriques Acked-by: Gregory Haskins Acked-by: Peter Zijlstra LKML-Reference: <20090324221002.GA10061@hades.domain.com> Signed-off-by: Ingo Molnar --- kernel/sched.c | 3 --- kernel/sched_debug.c | 5 +---- kernel/sched_stats.h | 7 +++---- 3 files changed, 4 insertions(+), 11 deletions(-) diff --git a/kernel/sched.c b/kernel/sched.c index d2dfe4c1a225..7b389c74f8ff 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -638,9 +638,6 @@ struct rq { /* could above be rq->cfs_rq.exec_clock + rq->rt_rq.rt_runtime ? */ /* sys_sched_yield() stats */ - unsigned int yld_exp_empty; - unsigned int yld_act_empty; - unsigned int yld_both_empty; unsigned int yld_count; /* schedule() stats */ diff --git a/kernel/sched_debug.c b/kernel/sched_debug.c index 4daebffa0565..467ca72f1657 100644 --- a/kernel/sched_debug.c +++ b/kernel/sched_debug.c @@ -286,9 +286,6 @@ static void print_cpu(struct seq_file *m, int cpu) #ifdef CONFIG_SCHEDSTATS #define P(n) SEQ_printf(m, " .%-30s: %d\n", #n, rq->n); - P(yld_exp_empty); - P(yld_act_empty); - P(yld_both_empty); P(yld_count); P(sched_switch); @@ -313,7 +310,7 @@ static int sched_debug_show(struct seq_file *m, void *v) u64 now = ktime_to_ns(ktime_get()); int cpu; - SEQ_printf(m, "Sched Debug Version: v0.08, %s %.*s\n", + SEQ_printf(m, "Sched Debug Version: v0.09, %s %.*s\n", init_utsname()->release, (int)strcspn(init_utsname()->version, " "), init_utsname()->version); diff --git a/kernel/sched_stats.h b/kernel/sched_stats.h index a8f93dd374e1..32d2bd4061b0 100644 --- a/kernel/sched_stats.h +++ b/kernel/sched_stats.h @@ -4,7 +4,7 @@ * bump this up when changing the output format or the meaning of an existing * format, so that tools can adapt (or abort) */ -#define SCHEDSTAT_VERSION 14 +#define SCHEDSTAT_VERSION 15 static int show_schedstat(struct seq_file *seq, void *v) { @@ -26,9 +26,8 @@ static int show_schedstat(struct seq_file *seq, void *v) /* runqueue-specific stats */ seq_printf(seq, - "cpu%d %u %u %u %u %u %u %u %u %u %llu %llu %lu", - cpu, rq->yld_both_empty, - rq->yld_act_empty, rq->yld_exp_empty, rq->yld_count, + "cpu%d %u %u %u %u %u %u %llu %llu %lu", + cpu, rq->yld_count, rq->sched_switch, rq->sched_count, rq->sched_goidle, rq->ttwu_count, rq->ttwu_local, rq->rq_cpu_time, From 67bb6c036d1fc3d332c8527a36a546e3e72e822c Mon Sep 17 00:00:00 2001 From: Gautham R Shenoy Date: Wed, 25 Mar 2009 14:43:35 +0530 Subject: [PATCH 36/46] sched: Simple helper functions for find_busiest_group() Impact: cleanup Currently the load idx calculation code is in find_busiest_group(). Move that to a static inline helper function. Similary, to find the first cpu of a sched_group we use cpumask_first(sched_group_cpus(group)) Use a helper to that. It improves readability in some cases. Signed-off-by: Gautham R Shenoy Acked-by: Peter Zijlstra Cc: Suresh Siddha Cc: "Balbir Singh" Cc: Nick Piggin Cc: "Dhaval Giani" Cc: Bharata B Rao Cc: "Vaidyanathan Srinivasan" LKML-Reference: <20090325091335.13992.55424.stgit@sofia.in.ibm.com> Signed-off-by: Ingo Molnar --- kernel/sched.c | 55 +++++++++++++++++++++++++++++++++++++++----------- 1 file changed, 43 insertions(+), 12 deletions(-) diff --git a/kernel/sched.c b/kernel/sched.c index 7b389c74f8ff..6aec1e7a72a3 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -3189,6 +3189,43 @@ static int move_one_task(struct rq *this_rq, int this_cpu, struct rq *busiest, return 0; } +/********** Helpers for find_busiest_group ************************/ + +/** + * group_first_cpu - Returns the first cpu in the cpumask of a sched_group. + * @group: The group whose first cpu is to be returned. + */ +static inline unsigned int group_first_cpu(struct sched_group *group) +{ + return cpumask_first(sched_group_cpus(group)); +} + +/** + * get_sd_load_idx - Obtain the load index for a given sched domain. + * @sd: The sched_domain whose load_idx is to be obtained. + * @idle: The Idle status of the CPU for whose sd load_icx is obtained. + */ +static inline int get_sd_load_idx(struct sched_domain *sd, + enum cpu_idle_type idle) +{ + int load_idx; + + switch (idle) { + case CPU_NOT_IDLE: + load_idx = sd->busy_idx; + break; + + case CPU_NEWLY_IDLE: + load_idx = sd->newidle_idx; + break; + default: + load_idx = sd->idle_idx; + break; + } + + return load_idx; +} +/******* find_busiest_group() helpers end here *********************/ /* * find_busiest_group finds and returns the busiest CPU group within the @@ -3217,12 +3254,7 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, busiest_load_per_task = busiest_nr_running = 0; this_load_per_task = this_nr_running = 0; - if (idle == CPU_NOT_IDLE) - load_idx = sd->busy_idx; - else if (idle == CPU_NEWLY_IDLE) - load_idx = sd->newidle_idx; - else - load_idx = sd->idle_idx; + load_idx = get_sd_load_idx(sd, idle); do { unsigned long load, group_capacity, max_cpu_load, min_cpu_load; @@ -3238,7 +3270,7 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, sched_group_cpus(group)); if (local_group) - balance_cpu = cpumask_first(sched_group_cpus(group)); + balance_cpu = group_first_cpu(group); /* Tally up the load of all CPUs in the group */ sum_weighted_load = sum_nr_running = avg_load = 0; @@ -3359,8 +3391,7 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, */ if ((sum_nr_running < min_nr_running) || (sum_nr_running == min_nr_running && - cpumask_first(sched_group_cpus(group)) > - cpumask_first(sched_group_cpus(group_min)))) { + group_first_cpu(group) > group_first_cpu(group_min))) { group_min = group; min_nr_running = sum_nr_running; min_load_per_task = sum_weighted_load / @@ -3375,8 +3406,8 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, if (sum_nr_running <= group_capacity - 1) { if (sum_nr_running > leader_nr_running || (sum_nr_running == leader_nr_running && - cpumask_first(sched_group_cpus(group)) < - cpumask_first(sched_group_cpus(group_leader)))) { + group_first_cpu(group) < + group_first_cpu(group_leader))) { group_leader = group; leader_nr_running = sum_nr_running; } @@ -3504,7 +3535,7 @@ out_balanced: *imbalance = min_load_per_task; if (sched_mc_power_savings >= POWERSAVINGS_BALANCE_WAKEUP) { cpu_rq(this_cpu)->rd->sched_mc_preferred_wakeup_cpu = - cpumask_first(sched_group_cpus(group_leader)); + group_first_cpu(group_leader); } return group_min; } From 6dfdb0629019f307ab18864b1fd3e5dbb02f383c Mon Sep 17 00:00:00 2001 From: Gautham R Shenoy Date: Wed, 25 Mar 2009 14:43:40 +0530 Subject: [PATCH 37/46] sched: Fix indentations in find_busiest_group() using gotos Impact: cleanup Some indentations in find_busiest_group() can minimized by using early exits with the help of gotos. This improves readability in a couple of cases. Signed-off-by: Gautham R Shenoy Acked-by: Peter Zijlstra Cc: Suresh Siddha Cc: "Balbir Singh" Cc: Nick Piggin Cc: "Dhaval Giani" Cc: Bharata B Rao Cc: "Vaidyanathan Srinivasan" LKML-Reference: <20090325091340.13992.45062.stgit@sofia.in.ibm.com> Signed-off-by: Ingo Molnar --- kernel/sched.c | 32 +++++++++++++++++--------------- 1 file changed, 17 insertions(+), 15 deletions(-) diff --git a/kernel/sched.c b/kernel/sched.c index 6aec1e7a72a3..f87adbe999e0 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -3403,14 +3403,14 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, * capacity but still has some space to pick up some load * from other group and save more power */ - if (sum_nr_running <= group_capacity - 1) { - if (sum_nr_running > leader_nr_running || - (sum_nr_running == leader_nr_running && - group_first_cpu(group) < - group_first_cpu(group_leader))) { - group_leader = group; - leader_nr_running = sum_nr_running; - } + if (sum_nr_running > group_capacity - 1) + goto group_next; + + if (sum_nr_running > leader_nr_running || + (sum_nr_running == leader_nr_running && + group_first_cpu(group) < group_first_cpu(group_leader))) { + group_leader = group; + leader_nr_running = sum_nr_running; } group_next: #endif @@ -3531,14 +3531,16 @@ out_balanced: if (idle == CPU_NOT_IDLE || !(sd->flags & SD_POWERSAVINGS_BALANCE)) goto ret; - if (this == group_leader && group_leader != group_min) { - *imbalance = min_load_per_task; - if (sched_mc_power_savings >= POWERSAVINGS_BALANCE_WAKEUP) { - cpu_rq(this_cpu)->rd->sched_mc_preferred_wakeup_cpu = - group_first_cpu(group_leader); - } - return group_min; + if (this != group_leader || group_leader == group_min) + goto ret; + + *imbalance = min_load_per_task; + if (sched_mc_power_savings >= POWERSAVINGS_BALANCE_WAKEUP) { + cpu_rq(this_cpu)->rd->sched_mc_preferred_wakeup_cpu = + group_first_cpu(group_leader); } + return group_min; + #endif ret: *imbalance = 0; From 381be78fdc829a22f6327a0ed09f54b6270a976d Mon Sep 17 00:00:00 2001 From: Gautham R Shenoy Date: Wed, 25 Mar 2009 14:43:46 +0530 Subject: [PATCH 38/46] sched: Define structure to store the sched_group statistics for fbg() Impact: cleanup Currently a whole bunch of variables are used to store the various statistics pertaining to the groups we iterate over in find_busiest_group(). Group them together in a single data structure and add appropriate comments. This will be useful later on when we create helper functions to calculate the sched_group statistics. Credit: Vaidyanathan Srinivasan Signed-off-by: Gautham R Shenoy Acked-by: Peter Zijlstra Cc: Suresh Siddha Cc: "Balbir Singh" Cc: Nick Piggin Cc: "Dhaval Giani" Cc: Bharata B Rao LKML-Reference: <20090325091345.13992.20099.stgit@sofia.in.ibm.com> Signed-off-by: Ingo Molnar --- kernel/sched.c | 79 +++++++++++++++++++++++++++++--------------------- 1 file changed, 46 insertions(+), 33 deletions(-) diff --git a/kernel/sched.c b/kernel/sched.c index f87adbe999e0..109db122de50 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -3191,6 +3191,18 @@ static int move_one_task(struct rq *this_rq, int this_cpu, struct rq *busiest, } /********** Helpers for find_busiest_group ************************/ +/** + * sg_lb_stats - stats of a sched_group required for load_balancing + */ +struct sg_lb_stats { + unsigned long avg_load; /*Avg load across the CPUs of the group */ + unsigned long group_load; /* Total load over the CPUs of the group */ + unsigned long sum_nr_running; /* Nr tasks running in the group */ + unsigned long sum_weighted_load; /* Weighted load of group's tasks */ + unsigned long group_capacity; + int group_imb; /* Is there an imbalance in the group ? */ +}; + /** * group_first_cpu - Returns the first cpu in the cpumask of a sched_group. * @group: The group whose first cpu is to be returned. @@ -3257,23 +3269,22 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, load_idx = get_sd_load_idx(sd, idle); do { - unsigned long load, group_capacity, max_cpu_load, min_cpu_load; + struct sg_lb_stats sgs; + unsigned long load, max_cpu_load, min_cpu_load; int local_group; int i; - int __group_imb = 0; unsigned int balance_cpu = -1, first_idle_cpu = 0; - unsigned long sum_nr_running, sum_weighted_load; unsigned long sum_avg_load_per_task; unsigned long avg_load_per_task; local_group = cpumask_test_cpu(this_cpu, sched_group_cpus(group)); + memset(&sgs, 0, sizeof(sgs)); if (local_group) balance_cpu = group_first_cpu(group); /* Tally up the load of all CPUs in the group */ - sum_weighted_load = sum_nr_running = avg_load = 0; sum_avg_load_per_task = avg_load_per_task = 0; max_cpu_load = 0; @@ -3301,9 +3312,9 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, min_cpu_load = load; } - avg_load += load; - sum_nr_running += rq->nr_running; - sum_weighted_load += weighted_cpuload(i); + sgs.group_load += load; + sgs.sum_nr_running += rq->nr_running; + sgs.sum_weighted_load += weighted_cpuload(i); sum_avg_load_per_task += cpu_avg_load_per_task(i); } @@ -3320,12 +3331,12 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, goto ret; } - total_load += avg_load; + total_load += sgs.group_load; total_pwr += group->__cpu_power; /* Adjust by relative CPU power of the group */ - avg_load = sg_div_cpu_power(group, - avg_load * SCHED_LOAD_SCALE); + sgs.avg_load = sg_div_cpu_power(group, + sgs.group_load * SCHED_LOAD_SCALE); /* @@ -3341,22 +3352,23 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, sum_avg_load_per_task * SCHED_LOAD_SCALE); if ((max_cpu_load - min_cpu_load) > 2*avg_load_per_task) - __group_imb = 1; + sgs.group_imb = 1; - group_capacity = group->__cpu_power / SCHED_LOAD_SCALE; + sgs.group_capacity = group->__cpu_power / SCHED_LOAD_SCALE; if (local_group) { - this_load = avg_load; + this_load = sgs.avg_load; this = group; - this_nr_running = sum_nr_running; - this_load_per_task = sum_weighted_load; - } else if (avg_load > max_load && - (sum_nr_running > group_capacity || __group_imb)) { - max_load = avg_load; + this_nr_running = sgs.sum_nr_running; + this_load_per_task = sgs.sum_weighted_load; + } else if (sgs.avg_load > max_load && + (sgs.sum_nr_running > sgs.group_capacity || + sgs.group_imb)) { + max_load = sgs.avg_load; busiest = group; - busiest_nr_running = sum_nr_running; - busiest_load_per_task = sum_weighted_load; - group_imb = __group_imb; + busiest_nr_running = sgs.sum_nr_running; + busiest_load_per_task = sgs.sum_weighted_load; + group_imb = sgs.group_imb; } #if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT) @@ -3372,7 +3384,7 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, * If the local group is idle or completely loaded * no need to do power savings balance at this domain */ - if (local_group && (this_nr_running >= group_capacity || + if (local_group && (this_nr_running >= sgs.group_capacity || !this_nr_running)) power_savings_balance = 0; @@ -3380,8 +3392,9 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, * If a group is already running at full capacity or idle, * don't include that group in power savings calculations */ - if (!power_savings_balance || sum_nr_running >= group_capacity - || !sum_nr_running) + if (!power_savings_balance || + sgs.sum_nr_running >= sgs.group_capacity || + !sgs.sum_nr_running) goto group_next; /* @@ -3389,13 +3402,13 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, * This is the group from where we need to pick up the load * for saving power */ - if ((sum_nr_running < min_nr_running) || - (sum_nr_running == min_nr_running && + if ((sgs.sum_nr_running < min_nr_running) || + (sgs.sum_nr_running == min_nr_running && group_first_cpu(group) > group_first_cpu(group_min))) { group_min = group; - min_nr_running = sum_nr_running; - min_load_per_task = sum_weighted_load / - sum_nr_running; + min_nr_running = sgs.sum_nr_running; + min_load_per_task = sgs.sum_weighted_load / + sgs.sum_nr_running; } /* @@ -3403,14 +3416,14 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, * capacity but still has some space to pick up some load * from other group and save more power */ - if (sum_nr_running > group_capacity - 1) + if (sgs.sum_nr_running > sgs.group_capacity - 1) goto group_next; - if (sum_nr_running > leader_nr_running || - (sum_nr_running == leader_nr_running && + if (sgs.sum_nr_running > leader_nr_running || + (sgs.sum_nr_running == leader_nr_running && group_first_cpu(group) < group_first_cpu(group_leader))) { group_leader = group; - leader_nr_running = sum_nr_running; + leader_nr_running = sgs.sum_nr_running; } group_next: #endif From 1f8c553d0f11d85f7993fe21015695d266771c00 Mon Sep 17 00:00:00 2001 From: Gautham R Shenoy Date: Wed, 25 Mar 2009 14:43:51 +0530 Subject: [PATCH 39/46] sched: Create a helper function to calculate sched_group stats for fbg() Impact: cleanup Create a helper function named update_sg_lb_stats() which can be invoked to calculate the individual group's statistics in find_busiest_group(). This reduces the lenght of find_busiest_group() considerably. Credit: Vaidyanathan Srinivasan Signed-off-by: Gautham R Shenoy Aked-by: Peter Zijlstra Cc: Suresh Siddha Cc: "Balbir Singh" Cc: Nick Piggin Cc: "Dhaval Giani" Cc: Bharata B Rao LKML-Reference: <20090325091351.13992.43461.stgit@sofia.in.ibm.com> Signed-off-by: Ingo Molnar --- kernel/sched.c | 175 ++++++++++++++++++++++++++++--------------------- 1 file changed, 100 insertions(+), 75 deletions(-) diff --git a/kernel/sched.c b/kernel/sched.c index 109db122de50..1893d5562f5f 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -3237,6 +3237,103 @@ static inline int get_sd_load_idx(struct sched_domain *sd, return load_idx; } + + +/** + * update_sg_lb_stats - Update sched_group's statistics for load balancing. + * @group: sched_group whose statistics are to be updated. + * @this_cpu: Cpu for which load balance is currently performed. + * @idle: Idle status of this_cpu + * @load_idx: Load index of sched_domain of this_cpu for load calc. + * @sd_idle: Idle status of the sched_domain containing group. + * @local_group: Does group contain this_cpu. + * @cpus: Set of cpus considered for load balancing. + * @balance: Should we balance. + * @sgs: variable to hold the statistics for this group. + */ +static inline void update_sg_lb_stats(struct sched_group *group, int this_cpu, + enum cpu_idle_type idle, int load_idx, int *sd_idle, + int local_group, const struct cpumask *cpus, + int *balance, struct sg_lb_stats *sgs) +{ + unsigned long load, max_cpu_load, min_cpu_load; + int i; + unsigned int balance_cpu = -1, first_idle_cpu = 0; + unsigned long sum_avg_load_per_task; + unsigned long avg_load_per_task; + + if (local_group) + balance_cpu = group_first_cpu(group); + + /* Tally up the load of all CPUs in the group */ + sum_avg_load_per_task = avg_load_per_task = 0; + max_cpu_load = 0; + min_cpu_load = ~0UL; + + for_each_cpu_and(i, sched_group_cpus(group), cpus) { + struct rq *rq = cpu_rq(i); + + if (*sd_idle && rq->nr_running) + *sd_idle = 0; + + /* Bias balancing toward cpus of our domain */ + if (local_group) { + if (idle_cpu(i) && !first_idle_cpu) { + first_idle_cpu = 1; + balance_cpu = i; + } + + load = target_load(i, load_idx); + } else { + load = source_load(i, load_idx); + if (load > max_cpu_load) + max_cpu_load = load; + if (min_cpu_load > load) + min_cpu_load = load; + } + + sgs->group_load += load; + sgs->sum_nr_running += rq->nr_running; + sgs->sum_weighted_load += weighted_cpuload(i); + + sum_avg_load_per_task += cpu_avg_load_per_task(i); + } + + /* + * First idle cpu or the first cpu(busiest) in this sched group + * is eligible for doing load balancing at this and above + * 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) { + *balance = 0; + return; + } + + /* Adjust by relative CPU power of the group */ + sgs->avg_load = sg_div_cpu_power(group, + sgs->group_load * SCHED_LOAD_SCALE); + + + /* + * Consider the group unbalanced when the imbalance is larger + * than the average weight of two tasks. + * + * APZ: with cgroup the avg task weight can vary wildly and + * might not be a suitable number - should we keep a + * normalized nr_running number somewhere that negates + * the hierarchy? + */ + avg_load_per_task = sg_div_cpu_power(group, + sum_avg_load_per_task * SCHED_LOAD_SCALE); + + if ((max_cpu_load - min_cpu_load) > 2*avg_load_per_task) + sgs->group_imb = 1; + + sgs->group_capacity = group->__cpu_power / SCHED_LOAD_SCALE; + +} /******* find_busiest_group() helpers end here *********************/ /* @@ -3270,92 +3367,20 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, do { struct sg_lb_stats sgs; - unsigned long load, max_cpu_load, min_cpu_load; int local_group; - int i; - unsigned int balance_cpu = -1, first_idle_cpu = 0; - unsigned long sum_avg_load_per_task; - unsigned long avg_load_per_task; local_group = cpumask_test_cpu(this_cpu, sched_group_cpus(group)); memset(&sgs, 0, sizeof(sgs)); + update_sg_lb_stats(group, this_cpu, idle, load_idx, sd_idle, + local_group, cpus, balance, &sgs); - if (local_group) - balance_cpu = group_first_cpu(group); - - /* Tally up the load of all CPUs in the group */ - sum_avg_load_per_task = avg_load_per_task = 0; - - max_cpu_load = 0; - min_cpu_load = ~0UL; - - for_each_cpu_and(i, sched_group_cpus(group), cpus) { - struct rq *rq = cpu_rq(i); - - if (*sd_idle && rq->nr_running) - *sd_idle = 0; - - /* Bias balancing toward cpus of our domain */ - if (local_group) { - if (idle_cpu(i) && !first_idle_cpu) { - first_idle_cpu = 1; - balance_cpu = i; - } - - load = target_load(i, load_idx); - } else { - load = source_load(i, load_idx); - if (load > max_cpu_load) - max_cpu_load = load; - if (min_cpu_load > load) - min_cpu_load = load; - } - - sgs.group_load += load; - sgs.sum_nr_running += rq->nr_running; - sgs.sum_weighted_load += weighted_cpuload(i); - - sum_avg_load_per_task += cpu_avg_load_per_task(i); - } - - /* - * First idle cpu or the first cpu(busiest) in this sched group - * is eligible for doing load balancing at this and above - * 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) { - *balance = 0; + if (balance && !(*balance)) goto ret; - } total_load += sgs.group_load; total_pwr += group->__cpu_power; - /* Adjust by relative CPU power of the group */ - sgs.avg_load = sg_div_cpu_power(group, - sgs.group_load * SCHED_LOAD_SCALE); - - - /* - * Consider the group unbalanced when the imbalance is larger - * than the average weight of two tasks. - * - * APZ: with cgroup the avg task weight can vary wildly and - * might not be a suitable number - should we keep a - * normalized nr_running number somewhere that negates - * the hierarchy? - */ - avg_load_per_task = sg_div_cpu_power(group, - sum_avg_load_per_task * SCHED_LOAD_SCALE); - - if ((max_cpu_load - min_cpu_load) > 2*avg_load_per_task) - sgs.group_imb = 1; - - sgs.group_capacity = group->__cpu_power / SCHED_LOAD_SCALE; - if (local_group) { this_load = sgs.avg_load; this = group; From 222d656dea57e4e084fbd1e9383e6fed2ca9fa61 Mon Sep 17 00:00:00 2001 From: Gautham R Shenoy Date: Wed, 25 Mar 2009 14:43:56 +0530 Subject: [PATCH 40/46] sched: Define structure to store the sched_domain statistics for fbg() Impact: cleanup Currently we use a lot of local variables in find_busiest_group() to capture the various statistics related to the sched_domain. Group them together into a single data structure. This will help us to offload the job of updating the sched_domain statistics to a helper function. Credit: Vaidyanathan Srinivasan Signed-off-by: Gautham R Shenoy Acked-by: Peter Zijlstra Cc: Suresh Siddha Cc: "Balbir Singh" Cc: Nick Piggin Cc: "Dhaval Giani" Cc: Bharata B Rao LKML-Reference: <20090325091356.13992.25970.stgit@sofia.in.ibm.com> Signed-off-by: Ingo Molnar --- kernel/sched.c | 205 +++++++++++++++++++++++++++++-------------------- 1 file changed, 120 insertions(+), 85 deletions(-) diff --git a/kernel/sched.c b/kernel/sched.c index 1893d5562f5f..8198dbe8e4aa 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -3190,6 +3190,37 @@ static int move_one_task(struct rq *this_rq, int this_cpu, struct rq *busiest, return 0; } /********** Helpers for find_busiest_group ************************/ +/** + * sd_lb_stats - Structure to store the statistics of a sched_domain + * during load balancing. + */ +struct sd_lb_stats { + struct sched_group *busiest; /* Busiest group in this sd */ + struct sched_group *this; /* Local group in this sd */ + unsigned long total_load; /* Total load of all groups in sd */ + unsigned long total_pwr; /* Total power of all groups in sd */ + unsigned long avg_load; /* Average load across all groups in sd */ + + /** Statistics of this group */ + unsigned long this_load; + unsigned long this_load_per_task; + unsigned long this_nr_running; + + /* Statistics of the busiest group */ + unsigned long max_load; + unsigned long busiest_load_per_task; + unsigned long busiest_nr_running; + + int group_imb; /* Is there imbalance in this sd */ +#if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT) + int power_savings_balance; /* Is powersave balance needed for this sd */ + struct sched_group *group_min; /* Least loaded group in sd */ + struct sched_group *group_leader; /* Group which relieves group_min */ + unsigned long min_load_per_task; /* load_per_task in group_min */ + unsigned long leader_nr_running; /* Nr running of group_leader */ + unsigned long min_nr_running; /* Nr running of group_min */ +#endif +}; /** * sg_lb_stats - stats of a sched_group required for load_balancing @@ -3346,23 +3377,16 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, unsigned long *imbalance, enum cpu_idle_type idle, int *sd_idle, const struct cpumask *cpus, int *balance) { - struct sched_group *busiest = NULL, *this = NULL, *group = sd->groups; - unsigned long max_load, avg_load, total_load, this_load, total_pwr; + struct sd_lb_stats sds; + struct sched_group *group = sd->groups; unsigned long max_pull; - unsigned long busiest_load_per_task, busiest_nr_running; - unsigned long this_load_per_task, this_nr_running; - int load_idx, group_imb = 0; + int load_idx; + + memset(&sds, 0, sizeof(sds)); #if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT) - int power_savings_balance = 1; - unsigned long leader_nr_running = 0, min_load_per_task = 0; - unsigned long min_nr_running = ULONG_MAX; - struct sched_group *group_min = NULL, *group_leader = NULL; + sds.power_savings_balance = 1; + sds.min_nr_running = ULONG_MAX; #endif - - max_load = this_load = total_load = total_pwr = 0; - busiest_load_per_task = busiest_nr_running = 0; - this_load_per_task = this_nr_running = 0; - load_idx = get_sd_load_idx(sd, idle); do { @@ -3378,22 +3402,22 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, if (balance && !(*balance)) goto ret; - total_load += sgs.group_load; - total_pwr += group->__cpu_power; + sds.total_load += sgs.group_load; + sds.total_pwr += group->__cpu_power; if (local_group) { - this_load = sgs.avg_load; - this = group; - this_nr_running = sgs.sum_nr_running; - this_load_per_task = sgs.sum_weighted_load; - } else if (sgs.avg_load > max_load && + sds.this_load = sgs.avg_load; + sds.this = group; + 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)) { - max_load = sgs.avg_load; - busiest = group; - busiest_nr_running = sgs.sum_nr_running; - busiest_load_per_task = sgs.sum_weighted_load; - group_imb = sgs.group_imb; + sds.max_load = sgs.avg_load; + sds.busiest = group; + sds.busiest_nr_running = sgs.sum_nr_running; + sds.busiest_load_per_task = sgs.sum_weighted_load; + sds.group_imb = sgs.group_imb; } #if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT) @@ -3409,15 +3433,16 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, * If the local group is idle or completely loaded * no need to do power savings balance at this domain */ - if (local_group && (this_nr_running >= sgs.group_capacity || - !this_nr_running)) - power_savings_balance = 0; + if (local_group && + (sds.this_nr_running >= sgs.group_capacity || + !sds.this_nr_running)) + sds.power_savings_balance = 0; /* * If a group is already running at full capacity or idle, * don't include that group in power savings calculations */ - if (!power_savings_balance || + if (!sds.power_savings_balance || sgs.sum_nr_running >= sgs.group_capacity || !sgs.sum_nr_running) goto group_next; @@ -3427,12 +3452,13 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, * This is the group from where we need to pick up the load * for saving power */ - if ((sgs.sum_nr_running < min_nr_running) || - (sgs.sum_nr_running == min_nr_running && - group_first_cpu(group) > group_first_cpu(group_min))) { - group_min = group; - min_nr_running = sgs.sum_nr_running; - min_load_per_task = sgs.sum_weighted_load / + if ((sgs.sum_nr_running < sds.min_nr_running) || + (sgs.sum_nr_running == sds.min_nr_running && + group_first_cpu(group) > + group_first_cpu(sds.group_min))) { + sds.group_min = group; + sds.min_nr_running = sgs.sum_nr_running; + sds.min_load_per_task = sgs.sum_weighted_load / sgs.sum_nr_running; } @@ -3444,29 +3470,32 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, if (sgs.sum_nr_running > sgs.group_capacity - 1) goto group_next; - if (sgs.sum_nr_running > leader_nr_running || - (sgs.sum_nr_running == leader_nr_running && - group_first_cpu(group) < group_first_cpu(group_leader))) { - group_leader = group; - leader_nr_running = sgs.sum_nr_running; + if (sgs.sum_nr_running > sds.leader_nr_running || + (sgs.sum_nr_running == sds.leader_nr_running && + group_first_cpu(group) < + group_first_cpu(sds.group_leader))) { + sds.group_leader = group; + sds.leader_nr_running = sgs.sum_nr_running; } group_next: #endif group = group->next; } while (group != sd->groups); - if (!busiest || this_load >= max_load || busiest_nr_running == 0) + if (!sds.busiest || sds.this_load >= sds.max_load + || sds.busiest_nr_running == 0) goto out_balanced; - avg_load = (SCHED_LOAD_SCALE * total_load) / total_pwr; + sds.avg_load = (SCHED_LOAD_SCALE * sds.total_load) / sds.total_pwr; - if (this_load >= avg_load || - 100*max_load <= sd->imbalance_pct*this_load) + if (sds.this_load >= sds.avg_load || + 100*sds.max_load <= sd->imbalance_pct * sds.this_load) goto out_balanced; - busiest_load_per_task /= busiest_nr_running; - if (group_imb) - busiest_load_per_task = min(busiest_load_per_task, avg_load); + sds.busiest_load_per_task /= sds.busiest_nr_running; + if (sds.group_imb) + sds.busiest_load_per_task = + min(sds.busiest_load_per_task, sds.avg_load); /* * We're trying to get all the cpus to the average_load, so we don't @@ -3479,7 +3508,7 @@ group_next: * by pulling tasks to us. Be careful of negative numbers as they'll * appear as very large values with unsigned longs. */ - if (max_load <= busiest_load_per_task) + if (sds.max_load <= sds.busiest_load_per_task) goto out_balanced; /* @@ -3487,17 +3516,18 @@ group_next: * max load less than avg load(as we skip the groups at or below * its cpu_power, while calculating max_load..) */ - if (max_load < avg_load) { + if (sds.max_load < sds.avg_load) { *imbalance = 0; goto small_imbalance; } /* Don't want to pull so many tasks that a group would go idle */ - max_pull = min(max_load - avg_load, max_load - busiest_load_per_task); + max_pull = min(sds.max_load - sds.avg_load, + sds.max_load - sds.busiest_load_per_task); /* How much load to actually move to equalise the imbalance */ - *imbalance = min(max_pull * busiest->__cpu_power, - (avg_load - this_load) * this->__cpu_power) + *imbalance = min(max_pull * sds.busiest->__cpu_power, + (sds.avg_load - sds.this_load) * sds.this->__cpu_power) / SCHED_LOAD_SCALE; /* @@ -3506,24 +3536,27 @@ group_next: * a think about bumping its value to force at least one task to be * moved */ - if (*imbalance < busiest_load_per_task) { + if (*imbalance < sds.busiest_load_per_task) { unsigned long tmp, pwr_now, pwr_move; unsigned int imbn; small_imbalance: pwr_move = pwr_now = 0; imbn = 2; - if (this_nr_running) { - this_load_per_task /= this_nr_running; - if (busiest_load_per_task > this_load_per_task) + if (sds.this_nr_running) { + sds.this_load_per_task /= sds.this_nr_running; + if (sds.busiest_load_per_task > + sds.this_load_per_task) imbn = 1; } else - this_load_per_task = cpu_avg_load_per_task(this_cpu); + sds.this_load_per_task = + cpu_avg_load_per_task(this_cpu); - if (max_load - this_load + busiest_load_per_task >= - busiest_load_per_task * imbn) { - *imbalance = busiest_load_per_task; - return busiest; + if (sds.max_load - sds.this_load + + sds.busiest_load_per_task >= + sds.busiest_load_per_task * imbn) { + *imbalance = sds.busiest_load_per_task; + return sds.busiest; } /* @@ -3532,52 +3565,54 @@ small_imbalance: * moving them. */ - pwr_now += busiest->__cpu_power * - min(busiest_load_per_task, max_load); - pwr_now += this->__cpu_power * - min(this_load_per_task, this_load); + pwr_now += sds.busiest->__cpu_power * + min(sds.busiest_load_per_task, sds.max_load); + pwr_now += sds.this->__cpu_power * + min(sds.this_load_per_task, sds.this_load); pwr_now /= SCHED_LOAD_SCALE; /* Amount of load we'd subtract */ - tmp = sg_div_cpu_power(busiest, - busiest_load_per_task * SCHED_LOAD_SCALE); - if (max_load > tmp) - pwr_move += busiest->__cpu_power * - min(busiest_load_per_task, max_load - tmp); + tmp = sg_div_cpu_power(sds.busiest, + sds.busiest_load_per_task * SCHED_LOAD_SCALE); + if (sds.max_load > tmp) + pwr_move += sds.busiest->__cpu_power * + min(sds.busiest_load_per_task, + sds.max_load - tmp); /* Amount of load we'd add */ - if (max_load * busiest->__cpu_power < - busiest_load_per_task * SCHED_LOAD_SCALE) - tmp = sg_div_cpu_power(this, - max_load * busiest->__cpu_power); + if (sds.max_load * sds.busiest->__cpu_power < + sds.busiest_load_per_task * SCHED_LOAD_SCALE) + tmp = sg_div_cpu_power(sds.this, + sds.max_load * sds.busiest->__cpu_power); else - tmp = sg_div_cpu_power(this, - busiest_load_per_task * SCHED_LOAD_SCALE); - pwr_move += this->__cpu_power * - min(this_load_per_task, this_load + tmp); + tmp = sg_div_cpu_power(sds.this, + sds.busiest_load_per_task * SCHED_LOAD_SCALE); + pwr_move += sds.this->__cpu_power * + min(sds.this_load_per_task, + sds.this_load + tmp); pwr_move /= SCHED_LOAD_SCALE; /* Move if we gain throughput */ if (pwr_move > pwr_now) - *imbalance = busiest_load_per_task; + *imbalance = sds.busiest_load_per_task; } - return busiest; + return sds.busiest; out_balanced: #if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT) if (idle == CPU_NOT_IDLE || !(sd->flags & SD_POWERSAVINGS_BALANCE)) goto ret; - if (this != group_leader || group_leader == group_min) + if (sds.this != sds.group_leader || sds.group_leader == sds.group_min) goto ret; - *imbalance = min_load_per_task; + *imbalance = sds.min_load_per_task; if (sched_mc_power_savings >= POWERSAVINGS_BALANCE_WAKEUP) { cpu_rq(this_cpu)->rd->sched_mc_preferred_wakeup_cpu = - group_first_cpu(group_leader); + group_first_cpu(sds.group_leader); } - return group_min; + return sds.group_min; #endif ret: From 37abe198b1246ddd206319c43502a687db62d347 Mon Sep 17 00:00:00 2001 From: Gautham R Shenoy Date: Wed, 25 Mar 2009 14:44:01 +0530 Subject: [PATCH 41/46] sched: Create a helper function to calculate sched_domain stats for fbg() Impact: cleanup Create a helper function named update_sd_lb_stats() to update the various sched_domain related statistics in find_busiest_group(). With this we would have moved all the statistics computation out of find_busiest_group(). Credit: Vaidyanathan Srinivasan Signed-off-by: Gautham R Shenoy Acked-by: Peter Zijlstra Cc: Suresh Siddha Cc: "Balbir Singh" Cc: Nick Piggin Cc: "Dhaval Giani" Cc: Bharata B Rao LKML-Reference: <20090325091401.13992.88737.stgit@sofia.in.ibm.com> Signed-off-by: Ingo Molnar --- kernel/sched.c | 117 ++++++++++++++++++++++++++++++------------------- 1 file changed, 73 insertions(+), 44 deletions(-) diff --git a/kernel/sched.c b/kernel/sched.c index 8198dbe8e4aa..ec715f97202e 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -3365,32 +3365,33 @@ static inline void update_sg_lb_stats(struct sched_group *group, int this_cpu, sgs->group_capacity = group->__cpu_power / SCHED_LOAD_SCALE; } -/******* find_busiest_group() helpers end here *********************/ -/* - * find_busiest_group finds and returns the busiest CPU group within the - * domain. It calculates and returns the amount of weighted load which - * should be moved to restore balance via the imbalance parameter. +/** + * update_sd_lb_stats - Update sched_group's statistics for load balancing. + * @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. + * @cpus: Set of cpus considered for load balancing. + * @balance: Should we balance. + * @sds: variable to hold the statistics for this sched_domain. */ -static struct sched_group * -find_busiest_group(struct sched_domain *sd, int this_cpu, - unsigned long *imbalance, enum cpu_idle_type idle, - int *sd_idle, const struct cpumask *cpus, int *balance) +static inline void update_sd_lb_stats(struct sched_domain *sd, int this_cpu, + enum cpu_idle_type idle, int *sd_idle, + const struct cpumask *cpus, int *balance, + struct sd_lb_stats *sds) { - struct sd_lb_stats sds; struct sched_group *group = sd->groups; - unsigned long max_pull; + struct sg_lb_stats sgs; int load_idx; - memset(&sds, 0, sizeof(sds)); #if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT) - sds.power_savings_balance = 1; - sds.min_nr_running = ULONG_MAX; + sds->power_savings_balance = 1; + sds->min_nr_running = ULONG_MAX; #endif load_idx = get_sd_load_idx(sd, idle); do { - struct sg_lb_stats sgs; int local_group; local_group = cpumask_test_cpu(this_cpu, @@ -3399,25 +3400,25 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, update_sg_lb_stats(group, this_cpu, idle, load_idx, sd_idle, local_group, cpus, balance, &sgs); - if (balance && !(*balance)) - goto ret; + if (local_group && balance && !(*balance)) + return; - sds.total_load += sgs.group_load; - sds.total_pwr += group->__cpu_power; + sds->total_load += sgs.group_load; + sds->total_pwr += group->__cpu_power; if (local_group) { - sds.this_load = sgs.avg_load; - sds.this = group; - 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 && + sds->this_load = sgs.avg_load; + sds->this = group; + 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)) { - sds.max_load = sgs.avg_load; - sds.busiest = group; - sds.busiest_nr_running = sgs.sum_nr_running; - sds.busiest_load_per_task = sgs.sum_weighted_load; - sds.group_imb = sgs.group_imb; + sds->max_load = sgs.avg_load; + sds->busiest = group; + sds->busiest_nr_running = sgs.sum_nr_running; + sds->busiest_load_per_task = sgs.sum_weighted_load; + sds->group_imb = sgs.group_imb; } #if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT) @@ -3434,15 +3435,15 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, * no need to do power savings balance at this domain */ if (local_group && - (sds.this_nr_running >= sgs.group_capacity || - !sds.this_nr_running)) - sds.power_savings_balance = 0; + (sds->this_nr_running >= sgs.group_capacity || + !sds->this_nr_running)) + sds->power_savings_balance = 0; /* * If a group is already running at full capacity or idle, * don't include that group in power savings calculations */ - if (!sds.power_savings_balance || + if (!sds->power_savings_balance || sgs.sum_nr_running >= sgs.group_capacity || !sgs.sum_nr_running) goto group_next; @@ -3452,13 +3453,13 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, * This is the group from where we need to pick up the load * for saving power */ - if ((sgs.sum_nr_running < sds.min_nr_running) || - (sgs.sum_nr_running == sds.min_nr_running && + if ((sgs.sum_nr_running < sds->min_nr_running) || + (sgs.sum_nr_running == sds->min_nr_running && group_first_cpu(group) > - group_first_cpu(sds.group_min))) { - sds.group_min = group; - sds.min_nr_running = sgs.sum_nr_running; - sds.min_load_per_task = sgs.sum_weighted_load / + group_first_cpu(sds->group_min))) { + sds->group_min = group; + sds->min_nr_running = sgs.sum_nr_running; + sds->min_load_per_task = sgs.sum_weighted_load / sgs.sum_nr_running; } @@ -3470,18 +3471,46 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, if (sgs.sum_nr_running > sgs.group_capacity - 1) goto group_next; - if (sgs.sum_nr_running > sds.leader_nr_running || - (sgs.sum_nr_running == sds.leader_nr_running && + if (sgs.sum_nr_running > sds->leader_nr_running || + (sgs.sum_nr_running == sds->leader_nr_running && group_first_cpu(group) < - group_first_cpu(sds.group_leader))) { - sds.group_leader = group; - sds.leader_nr_running = sgs.sum_nr_running; + group_first_cpu(sds->group_leader))) { + sds->group_leader = group; + sds->leader_nr_running = sgs.sum_nr_running; } group_next: #endif group = group->next; } while (group != sd->groups); +} +/******* find_busiest_group() helpers end here *********************/ + +/* + * find_busiest_group finds and returns the busiest CPU group within the + * domain. It calculates and returns the amount of weighted load which + * should be moved to restore balance via the imbalance parameter. + */ +static struct sched_group * +find_busiest_group(struct sched_domain *sd, int this_cpu, + unsigned long *imbalance, enum cpu_idle_type idle, + int *sd_idle, const struct cpumask *cpus, int *balance) +{ + struct sd_lb_stats sds; + unsigned long max_pull; + + memset(&sds, 0, sizeof(sds)); + + /* + * Compute the various statistics relavent for load balancing at + * this level. + */ + update_sd_lb_stats(sd, this_cpu, idle, sd_idle, cpus, + balance, &sds); + + if (balance && !(*balance)) + goto ret; + if (!sds.busiest || sds.this_load >= sds.max_load || sds.busiest_nr_running == 0) goto out_balanced; From 2e6f44aeda426054fc58464df1ad571aecca0c92 Mon Sep 17 00:00:00 2001 From: Gautham R Shenoy Date: Wed, 25 Mar 2009 14:44:06 +0530 Subject: [PATCH 42/46] sched: Create helper to calculate small_imbalance in fbg() Impact: cleanup We have two places in find_busiest_group() where we need to calculate the minor imbalance before returning the busiest group. Encapsulate this functionality into a seperate helper function. Credit: Vaidyanathan Srinivasan Signed-off-by: Gautham R Shenoy Acked-by: Peter Zijlstra Cc: Suresh Siddha Cc: "Balbir Singh" Cc: Nick Piggin Cc: "Dhaval Giani" Cc: Bharata B Rao LKML-Reference: <20090325091406.13992.54316.stgit@sofia.in.ibm.com> Signed-off-by: Ingo Molnar --- kernel/sched.c | 131 ++++++++++++++++++++++++++----------------------- 1 file changed, 70 insertions(+), 61 deletions(-) diff --git a/kernel/sched.c b/kernel/sched.c index ec715f97202e..540147e5e82b 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -3484,6 +3484,71 @@ group_next: } while (group != sd->groups); } + +/** + * fix_small_imbalance - Calculate the minor imbalance that exists + * amongst the groups of a sched_domain, during + * load balancing. + * @sds: Statistics of the sched_domain whose imbalance is to be calculated. + * @this_cpu: The cpu at whose sched_domain we're performing load-balance. + * @imbalance: Variable to store the imbalance. + */ +static inline void fix_small_imbalance(struct sd_lb_stats *sds, + int this_cpu, unsigned long *imbalance) +{ + unsigned long tmp, pwr_now = 0, pwr_move = 0; + unsigned int imbn = 2; + + if (sds->this_nr_running) { + sds->this_load_per_task /= sds->this_nr_running; + if (sds->busiest_load_per_task > + sds->this_load_per_task) + imbn = 1; + } else + sds->this_load_per_task = + cpu_avg_load_per_task(this_cpu); + + if (sds->max_load - sds->this_load + sds->busiest_load_per_task >= + sds->busiest_load_per_task * imbn) { + *imbalance = sds->busiest_load_per_task; + return; + } + + /* + * OK, we don't have enough imbalance to justify moving tasks, + * however we may be able to increase total CPU power used by + * moving them. + */ + + pwr_now += sds->busiest->__cpu_power * + min(sds->busiest_load_per_task, sds->max_load); + pwr_now += sds->this->__cpu_power * + min(sds->this_load_per_task, sds->this_load); + pwr_now /= SCHED_LOAD_SCALE; + + /* Amount of load we'd subtract */ + tmp = sg_div_cpu_power(sds->busiest, + sds->busiest_load_per_task * SCHED_LOAD_SCALE); + if (sds->max_load > tmp) + pwr_move += sds->busiest->__cpu_power * + min(sds->busiest_load_per_task, sds->max_load - tmp); + + /* Amount of load we'd add */ + if (sds->max_load * sds->busiest->__cpu_power < + sds->busiest_load_per_task * SCHED_LOAD_SCALE) + tmp = sg_div_cpu_power(sds->this, + sds->max_load * sds->busiest->__cpu_power); + else + tmp = sg_div_cpu_power(sds->this, + sds->busiest_load_per_task * SCHED_LOAD_SCALE); + pwr_move += sds->this->__cpu_power * + min(sds->this_load_per_task, sds->this_load + tmp); + pwr_move /= SCHED_LOAD_SCALE; + + /* Move if we gain throughput */ + if (pwr_move > pwr_now) + *imbalance = sds->busiest_load_per_task; +} /******* find_busiest_group() helpers end here *********************/ /* @@ -3547,7 +3612,8 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, */ if (sds.max_load < sds.avg_load) { *imbalance = 0; - goto small_imbalance; + fix_small_imbalance(&sds, this_cpu, imbalance); + goto ret_busiest; } /* Don't want to pull so many tasks that a group would go idle */ @@ -3565,67 +3631,10 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, * a think about bumping its value to force at least one task to be * moved */ - if (*imbalance < sds.busiest_load_per_task) { - unsigned long tmp, pwr_now, pwr_move; - unsigned int imbn; - -small_imbalance: - pwr_move = pwr_now = 0; - imbn = 2; - if (sds.this_nr_running) { - sds.this_load_per_task /= sds.this_nr_running; - if (sds.busiest_load_per_task > - sds.this_load_per_task) - imbn = 1; - } else - sds.this_load_per_task = - cpu_avg_load_per_task(this_cpu); - - if (sds.max_load - sds.this_load + - sds.busiest_load_per_task >= - sds.busiest_load_per_task * imbn) { - *imbalance = sds.busiest_load_per_task; - return sds.busiest; - } - - /* - * OK, we don't have enough imbalance to justify moving tasks, - * however we may be able to increase total CPU power used by - * moving them. - */ - - pwr_now += sds.busiest->__cpu_power * - min(sds.busiest_load_per_task, sds.max_load); - pwr_now += sds.this->__cpu_power * - min(sds.this_load_per_task, sds.this_load); - pwr_now /= SCHED_LOAD_SCALE; - - /* Amount of load we'd subtract */ - tmp = sg_div_cpu_power(sds.busiest, - sds.busiest_load_per_task * SCHED_LOAD_SCALE); - if (sds.max_load > tmp) - pwr_move += sds.busiest->__cpu_power * - min(sds.busiest_load_per_task, - sds.max_load - tmp); - - /* Amount of load we'd add */ - if (sds.max_load * sds.busiest->__cpu_power < - sds.busiest_load_per_task * SCHED_LOAD_SCALE) - tmp = sg_div_cpu_power(sds.this, - sds.max_load * sds.busiest->__cpu_power); - else - tmp = sg_div_cpu_power(sds.this, - sds.busiest_load_per_task * SCHED_LOAD_SCALE); - pwr_move += sds.this->__cpu_power * - min(sds.this_load_per_task, - sds.this_load + tmp); - pwr_move /= SCHED_LOAD_SCALE; - - /* Move if we gain throughput */ - if (pwr_move > pwr_now) - *imbalance = sds.busiest_load_per_task; - } + if (*imbalance < sds.busiest_load_per_task) + fix_small_imbalance(&sds, this_cpu, imbalance); +ret_busiest: return sds.busiest; out_balanced: From dbc523a3b86f9e1765b5e70e6886913b99cc5cec Mon Sep 17 00:00:00 2001 From: Gautham R Shenoy Date: Wed, 25 Mar 2009 14:44:12 +0530 Subject: [PATCH 43/46] sched: Create a helper function to calculate imbalance Move all the imbalance calculation out of find_busiest_group() through this helper function. With this change, the structure of find_busiest_group() will be as follows: - update_sched_domain_statistics. - check if imbalance exits. - update imbalance and return busiest. Signed-off-by: Gautham R Shenoy Acked-by: Peter Zijlstra Cc: Suresh Siddha Cc: "Balbir Singh" Cc: Nick Piggin Cc: "Dhaval Giani" Cc: Bharata B Rao Cc: "Vaidyanathan Srinivasan" LKML-Reference: <20090325091411.13992.43293.stgit@sofia.in.ibm.com> Signed-off-by: Ingo Molnar --- kernel/sched.c | 78 +++++++++++++++++++++++++++++--------------------- 1 file changed, 45 insertions(+), 33 deletions(-) diff --git a/kernel/sched.c b/kernel/sched.c index 540147e5e82b..934f615ccceb 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -3487,8 +3487,8 @@ group_next: /** * fix_small_imbalance - Calculate the minor imbalance that exists - * amongst the groups of a sched_domain, during - * load balancing. + * amongst the groups of a sched_domain, during + * load balancing. * @sds: Statistics of the sched_domain whose imbalance is to be calculated. * @this_cpu: The cpu at whose sched_domain we're performing load-balance. * @imbalance: Variable to store the imbalance. @@ -3549,6 +3549,47 @@ static inline void fix_small_imbalance(struct sd_lb_stats *sds, if (pwr_move > pwr_now) *imbalance = sds->busiest_load_per_task; } + +/** + * calculate_imbalance - Calculate the amount of imbalance present within the + * groups of a given sched_domain during load balance. + * @sds: statistics of the sched_domain whose imbalance is to be calculated. + * @this_cpu: Cpu for which currently load balance is being performed. + * @imbalance: The variable to store the imbalance. + */ +static inline void calculate_imbalance(struct sd_lb_stats *sds, int this_cpu, + unsigned long *imbalance) +{ + unsigned long max_pull; + /* + * 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_power, while calculating max_load..) + */ + if (sds->max_load < sds->avg_load) { + *imbalance = 0; + return fix_small_imbalance(sds, this_cpu, imbalance); + } + + /* Don't want to pull so many tasks that a group would go idle */ + max_pull = min(sds->max_load - sds->avg_load, + sds->max_load - sds->busiest_load_per_task); + + /* How much load to actually move to equalise the imbalance */ + *imbalance = min(max_pull * sds->busiest->__cpu_power, + (sds->avg_load - sds->this_load) * sds->this->__cpu_power) + / SCHED_LOAD_SCALE; + + /* + * if *imbalance is less than the average load per runnable task + * there is no gaurantee that any tasks will be moved so we'll have + * a think about bumping its value to force at least one task to be + * moved + */ + if (*imbalance < sds->busiest_load_per_task) + return fix_small_imbalance(sds, this_cpu, imbalance); + +} /******* find_busiest_group() helpers end here *********************/ /* @@ -3562,7 +3603,6 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, int *sd_idle, const struct cpumask *cpus, int *balance) { struct sd_lb_stats sds; - unsigned long max_pull; memset(&sds, 0, sizeof(sds)); @@ -3605,36 +3645,8 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, if (sds.max_load <= sds.busiest_load_per_task) goto out_balanced; - /* - * 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_power, while calculating max_load..) - */ - if (sds.max_load < sds.avg_load) { - *imbalance = 0; - fix_small_imbalance(&sds, this_cpu, imbalance); - goto ret_busiest; - } - - /* Don't want to pull so many tasks that a group would go idle */ - max_pull = min(sds.max_load - sds.avg_load, - sds.max_load - sds.busiest_load_per_task); - - /* How much load to actually move to equalise the imbalance */ - *imbalance = min(max_pull * sds.busiest->__cpu_power, - (sds.avg_load - sds.this_load) * sds.this->__cpu_power) - / SCHED_LOAD_SCALE; - - /* - * if *imbalance is less than the average load per runnable task - * there is no gaurantee that any tasks will be moved so we'll have - * a think about bumping its value to force at least one task to be - * moved - */ - if (*imbalance < sds.busiest_load_per_task) - fix_small_imbalance(&sds, this_cpu, imbalance); - -ret_busiest: + /* Looks like there is an imbalance. Compute it */ + calculate_imbalance(&sds, this_cpu, imbalance); return sds.busiest; out_balanced: From a021dc03376707c55a3483e32c16b8986d4414cc Mon Sep 17 00:00:00 2001 From: Gautham R Shenoy Date: Wed, 25 Mar 2009 14:44:17 +0530 Subject: [PATCH 44/46] sched: Optimize the !power_savings_balance during fbg() Impact: cleanup, micro-optimization We don't need to perform power_savings balance if either the cpu is NOT_IDLE or if the sched_domain doesn't contain the SD_POWERSAVINGS_BALANCE flag set. Currently, we check for these conditions multiple number of times, even though these variables don't change over the scope of find_busiest_group(). Check once, and store the value in the already exiting "power_savings_balance" variable. Signed-off-by: Gautham R Shenoy Acked-by: Peter Zijlstra Cc: Suresh Siddha Cc: "Balbir Singh" Cc: Nick Piggin Cc: "Dhaval Giani" Cc: Bharata B Rao Cc: "Vaidyanathan Srinivasan" LKML-Reference: <20090325091417.13992.2657.stgit@sofia.in.ibm.com> Signed-off-by: Ingo Molnar --- kernel/sched.c | 23 ++++++++++++++--------- 1 file changed, 14 insertions(+), 9 deletions(-) diff --git a/kernel/sched.c b/kernel/sched.c index 934f615ccceb..71e8dcaf2c79 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -3386,8 +3386,17 @@ static inline void update_sd_lb_stats(struct sched_domain *sd, int this_cpu, int load_idx; #if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT) - sds->power_savings_balance = 1; - sds->min_nr_running = ULONG_MAX; + /* + * Busy processors will not participate in power savings + * balance. + */ + if (idle == CPU_NOT_IDLE || !(sd->flags & SD_POWERSAVINGS_BALANCE)) + sds->power_savings_balance = 0; + else { + sds->power_savings_balance = 1; + sds->min_nr_running = ULONG_MAX; + sds->leader_nr_running = 0; + } #endif load_idx = get_sd_load_idx(sd, idle); @@ -3422,12 +3431,8 @@ static inline void update_sd_lb_stats(struct sched_domain *sd, int this_cpu, } #if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT) - /* - * Busy processors will not participate in power savings - * balance. - */ - if (idle == CPU_NOT_IDLE || - !(sd->flags & SD_POWERSAVINGS_BALANCE)) + + if (!sds->power_savings_balance) goto group_next; /* @@ -3651,7 +3656,7 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, out_balanced: #if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT) - if (idle == CPU_NOT_IDLE || !(sd->flags & SD_POWERSAVINGS_BALANCE)) + if (!sds.power_savings_balance) goto ret; if (sds.this != sds.group_leader || sds.group_leader == sds.group_min) From c071df18525a95b37dd5821a6dc4af83bd18675e Mon Sep 17 00:00:00 2001 From: Gautham R Shenoy Date: Wed, 25 Mar 2009 14:44:22 +0530 Subject: [PATCH 45/46] sched: Refactor the power savings balance code Impact: cleanup Create seperate helper functions to initialize the power-savings-balance related variables, to update them and to check if we have a scope for performing power-savings balance. Add no-op inline functions for the !(CONFIG_SCHED_MC || CONFIG_SCHED_SMT) case. This will eliminate all the #ifdef jungle in find_busiest_group() and the other helper functions. Signed-off-by: Gautham R Shenoy Acked-by: Peter Zijlstra Cc: Suresh Siddha Cc: "Balbir Singh" Cc: Nick Piggin Cc: "Dhaval Giani" Cc: Bharata B Rao Cc: "Vaidyanathan Srinivasan" LKML-Reference: <20090325091422.13992.73616.stgit@sofia.in.ibm.com> Signed-off-by: Ingo Molnar --- kernel/sched.c | 236 ++++++++++++++++++++++++++++++++----------------- 1 file changed, 153 insertions(+), 83 deletions(-) diff --git a/kernel/sched.c b/kernel/sched.c index 71e8dcaf2c79..5f21658b0f67 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -3270,6 +3270,151 @@ static inline int get_sd_load_idx(struct sched_domain *sd, } +#if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT) +/** + * init_sd_power_savings_stats - Initialize power savings statistics for + * the given sched_domain, during load balancing. + * + * @sd: Sched domain whose power-savings statistics are to be initialized. + * @sds: Variable containing the statistics for sd. + * @idle: Idle status of the CPU at which we're performing load-balancing. + */ +static inline void init_sd_power_savings_stats(struct sched_domain *sd, + struct sd_lb_stats *sds, enum cpu_idle_type idle) +{ + /* + * Busy processors will not participate in power savings + * balance. + */ + if (idle == CPU_NOT_IDLE || !(sd->flags & SD_POWERSAVINGS_BALANCE)) + sds->power_savings_balance = 0; + else { + sds->power_savings_balance = 1; + sds->min_nr_running = ULONG_MAX; + sds->leader_nr_running = 0; + } +} + +/** + * update_sd_power_savings_stats - Update the power saving stats for a + * sched_domain while performing load balancing. + * + * @group: sched_group belonging to the sched_domain under consideration. + * @sds: Variable containing the statistics of the sched_domain + * @local_group: Does group contain the CPU for which we're performing + * load balancing ? + * @sgs: Variable containing the statistics of the group. + */ +static inline void update_sd_power_savings_stats(struct sched_group *group, + struct sd_lb_stats *sds, int local_group, struct sg_lb_stats *sgs) +{ + + if (!sds->power_savings_balance) + return; + + /* + * If the local group is idle or completely loaded + * no need to do power savings balance at this domain + */ + if (local_group && (sds->this_nr_running >= sgs->group_capacity || + !sds->this_nr_running)) + sds->power_savings_balance = 0; + + /* + * If a group is already running at full capacity or idle, + * don't include that group in power savings calculations + */ + if (!sds->power_savings_balance || + sgs->sum_nr_running >= sgs->group_capacity || + !sgs->sum_nr_running) + return; + + /* + * Calculate the group which has the least non-idle load. + * This is the group from where we need to pick up the load + * for saving power + */ + if ((sgs->sum_nr_running < sds->min_nr_running) || + (sgs->sum_nr_running == sds->min_nr_running && + group_first_cpu(group) > group_first_cpu(sds->group_min))) { + sds->group_min = group; + sds->min_nr_running = sgs->sum_nr_running; + sds->min_load_per_task = sgs->sum_weighted_load / + sgs->sum_nr_running; + } + + /* + * Calculate the group which is almost near its + * capacity but still has some space to pick up some load + * from other group and save more power + */ + if (sgs->sum_nr_running > sgs->group_capacity - 1) + return; + + if (sgs->sum_nr_running > sds->leader_nr_running || + (sgs->sum_nr_running == sds->leader_nr_running && + group_first_cpu(group) < group_first_cpu(sds->group_leader))) { + sds->group_leader = group; + sds->leader_nr_running = sgs->sum_nr_running; + } +} + +/** + * check_power_save_busiest_group - Check if we have potential to perform + * some power-savings balance. If yes, set the busiest group to be + * the least loaded group in the sched_domain, so that it's CPUs can + * be put to idle. + * + * @sds: Variable containing the statistics of the sched_domain + * under consideration. + * @this_cpu: Cpu at which we're currently performing load-balancing. + * @imbalance: Variable to store the imbalance. + * + * Returns 1 if there is potential to perform power-savings balance. + * Else returns 0. + */ +static inline int check_power_save_busiest_group(struct sd_lb_stats *sds, + int this_cpu, unsigned long *imbalance) +{ + if (!sds->power_savings_balance) + return 0; + + if (sds->this != sds->group_leader || + sds->group_leader == sds->group_min) + return 0; + + *imbalance = sds->min_load_per_task; + sds->busiest = sds->group_min; + + if (sched_mc_power_savings >= POWERSAVINGS_BALANCE_WAKEUP) { + cpu_rq(this_cpu)->rd->sched_mc_preferred_wakeup_cpu = + group_first_cpu(sds->group_leader); + } + + return 1; + +} +#else /* CONFIG_SCHED_MC || CONFIG_SCHED_SMT */ +static inline void init_sd_power_savings_stats(struct sched_domain *sd, + struct sd_lb_stats *sds, enum cpu_idle_type idle) +{ + return; +} + +static inline void update_sd_power_savings_stats(struct sched_group *group, + struct sd_lb_stats *sds, int local_group, struct sg_lb_stats *sgs) +{ + return; +} + +static inline int check_power_save_busiest_group(struct sd_lb_stats *sds, + int this_cpu, unsigned long *imbalance) +{ + return 0; +} +#endif /* CONFIG_SCHED_MC || CONFIG_SCHED_SMT */ + + /** * update_sg_lb_stats - Update sched_group's statistics for load balancing. * @group: sched_group whose statistics are to be updated. @@ -3385,19 +3530,7 @@ static inline void update_sd_lb_stats(struct sched_domain *sd, int this_cpu, struct sg_lb_stats sgs; int load_idx; -#if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT) - /* - * Busy processors will not participate in power savings - * balance. - */ - if (idle == CPU_NOT_IDLE || !(sd->flags & SD_POWERSAVINGS_BALANCE)) - sds->power_savings_balance = 0; - else { - sds->power_savings_balance = 1; - sds->min_nr_running = ULONG_MAX; - sds->leader_nr_running = 0; - } -#endif + init_sd_power_savings_stats(sd, sds, idle); load_idx = get_sd_load_idx(sd, idle); do { @@ -3430,61 +3563,7 @@ static inline void update_sd_lb_stats(struct sched_domain *sd, int this_cpu, sds->group_imb = sgs.group_imb; } -#if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT) - - if (!sds->power_savings_balance) - goto group_next; - - /* - * If the local group is idle or completely loaded - * no need to do power savings balance at this domain - */ - if (local_group && - (sds->this_nr_running >= sgs.group_capacity || - !sds->this_nr_running)) - sds->power_savings_balance = 0; - - /* - * If a group is already running at full capacity or idle, - * don't include that group in power savings calculations - */ - if (!sds->power_savings_balance || - sgs.sum_nr_running >= sgs.group_capacity || - !sgs.sum_nr_running) - goto group_next; - - /* - * Calculate the group which has the least non-idle load. - * This is the group from where we need to pick up the load - * for saving power - */ - if ((sgs.sum_nr_running < sds->min_nr_running) || - (sgs.sum_nr_running == sds->min_nr_running && - group_first_cpu(group) > - group_first_cpu(sds->group_min))) { - sds->group_min = group; - sds->min_nr_running = sgs.sum_nr_running; - sds->min_load_per_task = sgs.sum_weighted_load / - sgs.sum_nr_running; - } - - /* - * Calculate the group which is almost near its - * capacity but still has some space to pick up some load - * from other group and save more power - */ - if (sgs.sum_nr_running > sgs.group_capacity - 1) - goto group_next; - - if (sgs.sum_nr_running > sds->leader_nr_running || - (sgs.sum_nr_running == sds->leader_nr_running && - group_first_cpu(group) < - group_first_cpu(sds->group_leader))) { - sds->group_leader = group; - sds->leader_nr_running = sgs.sum_nr_running; - } -group_next: -#endif + update_sd_power_savings_stats(group, sds, local_group, &sgs); group = group->next; } while (group != sd->groups); @@ -3655,21 +3734,12 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, return sds.busiest; out_balanced: -#if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT) - if (!sds.power_savings_balance) - goto ret; - - if (sds.this != sds.group_leader || sds.group_leader == sds.group_min) - goto ret; - - *imbalance = sds.min_load_per_task; - if (sched_mc_power_savings >= POWERSAVINGS_BALANCE_WAKEUP) { - cpu_rq(this_cpu)->rd->sched_mc_preferred_wakeup_cpu = - group_first_cpu(sds.group_leader); - } - return sds.group_min; - -#endif + /* + * There is no obvious imbalance. But check if we can do some balancing + * to save power. + */ + if (check_power_save_busiest_group(&sds, this_cpu, imbalance)) + return sds.busiest; ret: *imbalance = 0; return NULL; From b7bb4c9bb01941fe8feb653f3410e7ed0c9bb786 Mon Sep 17 00:00:00 2001 From: Gautham R Shenoy Date: Wed, 25 Mar 2009 14:44:27 +0530 Subject: [PATCH 46/46] sched: Add comments to find_busiest_group() function Impact: cleanup Add /** style comments around find_busiest_group(). Also add a few explanatory comments. This concludes the find_busiest_group() cleanup. The function is now down to 72 lines from the original 313 lines. Signed-off-by: Gautham R Shenoy Acked-by: Peter Zijlstra Cc: Suresh Siddha Cc: "Balbir Singh" Cc: Nick Piggin Cc: "Dhaval Giani" Cc: Bharata B Rao Cc: "Vaidyanathan Srinivasan" LKML-Reference: <20090325091427.13992.18933.stgit@sofia.in.ibm.com> Signed-off-by: Ingo Molnar --- kernel/sched.c | 50 ++++++++++++++++++++++++++++++++++++++++++-------- 1 file changed, 42 insertions(+), 8 deletions(-) diff --git a/kernel/sched.c b/kernel/sched.c index 5f21658b0f67..9f8506d68fdc 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -3676,10 +3676,30 @@ static inline void calculate_imbalance(struct sd_lb_stats *sds, int this_cpu, } /******* find_busiest_group() helpers end here *********************/ -/* - * find_busiest_group finds and returns the busiest CPU group within the - * domain. It calculates and returns the amount of weighted load which - * should be moved to restore balance via the imbalance parameter. +/** + * 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. + * + * Also calculates the amount of weighted load which should be moved + * to restore balance. + * + * @sd: The sched_domain whose busiest group is to be returned. + * @this_cpu: The cpu for which load balancing is currently being performed. + * @imbalance: Variable which stores amount of weighted load which should + * be moved to restore balance/put a group to idle. + * @idle: The idle status of this_cpu. + * @sd_idle: The idleness of sd + * @cpus: The set of CPUs under consideration for load-balancing. + * @balance: Pointer to a variable indicating if this_cpu + * is the appropriate cpu to perform load balancing at this_level. + * + * Returns: - 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 sched_domain *sd, int this_cpu, @@ -3697,17 +3717,31 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, update_sd_lb_stats(sd, this_cpu, idle, sd_idle, cpus, balance, &sds); + /* Cases where imbalance does not exist from POV of this_cpu */ + /* 1) this_cpu is not the appropriate cpu to perform load balancing + * at this level. + * 2) There is no busy sibling group to pull from. + * 3) This group is the busiest group. + * 4) This group is more busy than the avg busieness at this + * sched_domain. + * 5) The imbalance is within the specified limit. + * 6) Any rebalance would lead to ping-pong + */ if (balance && !(*balance)) goto ret; - if (!sds.busiest || sds.this_load >= sds.max_load - || sds.busiest_nr_running == 0) + if (!sds.busiest || sds.busiest_nr_running == 0) + goto out_balanced; + + if (sds.this_load >= sds.max_load) goto out_balanced; sds.avg_load = (SCHED_LOAD_SCALE * sds.total_load) / sds.total_pwr; - if (sds.this_load >= sds.avg_load || - 100*sds.max_load <= sd->imbalance_pct * sds.this_load) + if (sds.this_load >= sds.avg_load) + goto out_balanced; + + if (100 * sds.max_load <= sd->imbalance_pct * sds.this_load) goto out_balanced; sds.busiest_load_per_task /= sds.busiest_nr_running;