diff -rup linux-2.6.7-rc3/arch/i386/Kconfig linux-2.6.7-rc3-ss/arch/i386/Kconfig --- linux-2.6.7-rc3/arch/i386/Kconfig Mon Jun 7 12:13:39 2004 +++ linux-2.6.7-rc3-ss/arch/i386/Kconfig Thu Jun 10 12:30:30 2004 @@ -1294,6 +1294,19 @@ config 4KSTACKS on the VM subsystem for higher order allocations. This option will also use IRQ stacks to compensate for the reduced stackspace. +config SCHEDSTATS + bool "Collect scheduler statistics" + depends on PROC_FS + default y + help + If you say Y here, additional code will be inserted into the + scheduler and related routines to collect statistics about + scheduler behavior and provide them in /proc/schedstat. These + stats may be useful for both tuning and debugging the scheduler + If you aren't debugging the scheduler or trying to tune a specific + application, you can say N to avoid the very slight overhead + this adds. + config X86_FIND_SMP_CONFIG bool depends on X86_LOCAL_APIC || X86_VOYAGER diff -rup linux-2.6.7-rc3/arch/ppc/Kconfig linux-2.6.7-rc3-ss/arch/ppc/Kconfig --- linux-2.6.7-rc3/arch/ppc/Kconfig Mon Jun 7 12:15:12 2004 +++ linux-2.6.7-rc3-ss/arch/ppc/Kconfig Thu Jun 10 12:30:30 2004 @@ -1251,6 +1251,19 @@ config DEBUG_INFO debug the kernel. If you don't debug the kernel, you can say N. +config SCHEDSTATS + bool "Collect scheduler statistics" + depends on PROC_FS + default y + help + If you say Y here, additional code will be inserted into the + scheduler and related routines to collect statistics about + scheduler behavior and provide them in /proc/schedstat. These + stats may be useful for both tuning and debugging the scheduler + If you aren't debugging the scheduler or trying to tune a specific + application, you can say N to avoid the very slight overhead + this adds. + config BOOTX_TEXT bool "Support for early boot text console (BootX or OpenFirmware only)" depends PPC_OF diff -rup linux-2.6.7-rc3/arch/ppc64/Kconfig linux-2.6.7-rc3-ss/arch/ppc64/Kconfig --- linux-2.6.7-rc3/arch/ppc64/Kconfig Mon Jun 7 12:14:25 2004 +++ linux-2.6.7-rc3-ss/arch/ppc64/Kconfig Thu Jun 10 12:30:30 2004 @@ -425,6 +425,19 @@ config IRQSTACKS for handling hard and soft interrupts. This can help avoid overflowing the process kernel stacks. +config SCHEDSTATS + bool "Collect scheduler statistics" + depends on PROC_FS + default y + help + If you say Y here, additional code will be inserted into the + scheduler and related routines to collect statistics about + scheduler behavior and provide them in /proc/schedstat. These + stats may be useful for both tuning and debugging the scheduler + If you aren't debugging the scheduler or trying to tune a specific + application, you can say N to avoid the very slight overhead + this adds. + endmenu config SPINLINE diff -rup linux-2.6.7-rc3/arch/x86_64/Kconfig linux-2.6.7-rc3-ss/arch/x86_64/Kconfig --- linux-2.6.7-rc3/arch/x86_64/Kconfig Mon Jun 7 12:14:56 2004 +++ linux-2.6.7-rc3-ss/arch/x86_64/Kconfig Thu Jun 10 12:30:30 2004 @@ -461,6 +461,19 @@ config DEBUG_INFO Say Y here only if you plan to use gdb to debug the kernel. Please note that this option requires new binutils. If you don't debug the kernel, you can say N. + +config SCHEDSTATS + bool "Collect scheduler statistics" + depends on PROC_FS + default y + help + If you say Y here, additional code will be inserted into the + scheduler and related routines to collect statistics about + scheduler behavior and provide them in /proc/schedstat. These + stats may be useful for both tuning and debugging the scheduler + If you aren't debugging the scheduler or trying to tune a specific + application, you can say N to avoid the very slight overhead + this adds. config FRAME_POINTER bool "Compile the kernel with frame pointers" diff -rup linux-2.6.7-rc3/fs/proc/array.c linux-2.6.7-rc3-ss/fs/proc/array.c --- linux-2.6.7-rc3/fs/proc/array.c Mon Jun 7 12:14:33 2004 +++ linux-2.6.7-rc3-ss/fs/proc/array.c Thu Jun 10 12:30:30 2004 @@ -357,9 +357,15 @@ int proc_pid_stat(struct task_struct *ta /* Temporary variable needed for gcc-2.96 */ start_time = jiffies_64_to_clock_t(task->start_time - INITIAL_JIFFIES); +#ifdef CONFIG_SCHEDSTATS + res = sprintf(buffer,"%d (%s) %c %d %d %d %d %d %lu %lu \ +%lu %lu %lu %lu %lu %ld %ld %ld %ld %d %ld %llu %lu %ld %lu %lu %lu %lu %lu \ +%lu %lu %lu %lu %lu %lu %lu %lu %d %d %lu %lu %lu %lu %lu\n", +#else res = sprintf(buffer,"%d (%s) %c %d %d %d %d %d %lu %lu \ %lu %lu %lu %lu %lu %ld %ld %ld %ld %d %ld %llu %lu %ld %lu %lu %lu %lu %lu \ %lu %lu %lu %lu %lu %lu %lu %lu %d %d %lu %lu\n", +#endif task->pid, task->comm, state, @@ -404,7 +410,14 @@ int proc_pid_stat(struct task_struct *ta task->exit_signal, task_cpu(task), task->rt_priority, +#ifdef CONFIG_SCHEDSTATS + task->policy, + task->sched_info.cpu_time, + task->sched_info.run_delay, + task->sched_info.pcnt); +#else task->policy); +#endif /* CONFIG_SCHEDSTATS */ if(mm) mmput(mm); return res; diff -rup linux-2.6.7-rc3/fs/proc/proc_misc.c linux-2.6.7-rc3-ss/fs/proc/proc_misc.c --- linux-2.6.7-rc3/fs/proc/proc_misc.c Mon Jun 7 12:13:39 2004 +++ linux-2.6.7-rc3-ss/fs/proc/proc_misc.c Thu Jun 10 12:30:30 2004 @@ -286,6 +286,10 @@ static struct file_operations proc_vmsta .release = seq_release, }; +#ifdef CONFIG_SCHEDSTATS +extern struct file_operations proc_schedstat_operations; +#endif + #ifdef CONFIG_PROC_HARDWARE static int hardware_read_proc(char *page, char **start, off_t off, int count, int *eof, void *data) @@ -697,6 +701,9 @@ void __init proc_misc_init(void) create_seq_entry("diskstats", 0, &proc_diskstats_operations); #ifdef CONFIG_MODULES create_seq_entry("modules", 0, &proc_modules_operations); +#endif +#ifdef CONFIG_SCHEDSTATS + create_seq_entry("schedstat", 0, &proc_schedstat_operations); #endif #ifdef CONFIG_PROC_KCORE proc_root_kcore = create_proc_entry("kcore", S_IRUSR, NULL); diff -rup linux-2.6.7-rc3/include/linux/sched.h linux-2.6.7-rc3-ss/include/linux/sched.h --- linux-2.6.7-rc3/include/linux/sched.h Mon Jun 7 12:13:38 2004 +++ linux-2.6.7-rc3-ss/include/linux/sched.h Fri Jun 11 17:16:38 2004 @@ -96,6 +96,16 @@ extern unsigned long nr_running(void); extern unsigned long nr_uninterruptible(void); extern unsigned long nr_iowait(void); +#ifdef CONFIG_SCHEDSTATS +struct sched_info; +extern void cpu_sched_info(struct sched_info *, int); +#define schedstat_inc(rq, field) rq->field++; +#define schedstat_add(rq, field, amt) rq->field += amt; +#else +#define schedstat_inc(rq, field) do { } while (0); +#define schedstat_add(rq, field, amt) do { } while (0); +#endif + #include #include #include @@ -347,6 +357,18 @@ struct k_itimer { struct sigqueue *sigq; /* signal queue entry. */ }; +#ifdef CONFIG_SCHEDSTATS +struct sched_info { + /* cumulative counters */ + unsigned long cpu_time, /* time spent on the cpu */ + run_delay, /* time spent waiting on a runqueue */ + pcnt; /* # of timeslices run on this cpu */ + + /* timestamps */ + unsigned long last_arrival, /* when we last ran on a cpu */ + last_queued; /* when we were last queued to run */ +}; +#endif /* CONFIG_SCHEDSTATS */ struct io_context; /* See blkdev.h */ void exit_io_context(void); @@ -403,6 +425,10 @@ struct task_struct { cpumask_t cpus_allowed; unsigned int time_slice, first_time_slice; +#ifdef CONFIG_SCHEDSTATS + struct sched_info sched_info; +#endif /* CONFIG_SCHEDSTATS */ + struct list_head tasks; struct list_head ptrace_children; struct list_head ptrace_list; @@ -561,6 +587,14 @@ do { if (atomic_dec_and_test(&(tsk)->usa #define SD_WAKE_BALANCE 32 /* Perform balancing at task wakeup */ #define SD_SHARE_CPUPOWER 64 /* Domain members share cpu power */ +enum idle_type +{ + IDLE, + NOT_IDLE, + NEWLY_IDLE, + MAX_IDLE_TYPES +}; + struct sched_group { struct sched_group *next; /* Must be a circular list */ cpumask_t cpumask; @@ -592,6 +626,23 @@ struct sched_domain { unsigned long last_balance; /* init to jiffies. units in jiffies */ unsigned int balance_interval; /* initialise to 1. units in ms. */ unsigned int nr_balance_failed; /* initialise to 0 */ + +#ifdef CONFIG_SCHEDSTATS + /* load_balance() stats */ + unsigned long lb_cnt[MAX_IDLE_TYPES]; + unsigned long lb_failed[MAX_IDLE_TYPES]; + unsigned long lb_imbalance[MAX_IDLE_TYPES]; + unsigned long lb_nobusyg[MAX_IDLE_TYPES]; + unsigned long lb_nobusyq[MAX_IDLE_TYPES]; + + /* sched_balance_exec() stats */ + unsigned long sbe_attempts; + unsigned long sbe_pushed; + + /* try_to_wake_up() stats */ + unsigned long ttwu_wake_affine; + unsigned long ttwu_wake_balance; +#endif }; /* Common values for SMT siblings */ diff -rup linux-2.6.7-rc3/kernel/fork.c linux-2.6.7-rc3-ss/kernel/fork.c --- linux-2.6.7-rc3/kernel/fork.c Mon Jun 7 12:13:38 2004 +++ linux-2.6.7-rc3-ss/kernel/fork.c Thu Jun 10 12:30:30 2004 @@ -964,6 +964,11 @@ struct task_struct *copy_process(unsigne p->security = NULL; p->io_context = NULL; p->audit_context = NULL; + +#ifdef CONFIG_SCHEDSTATS + memset(&p->sched_info, 0, sizeof(p->sched_info)); +#endif /* CONFIG_SCHEDSTATS */ + #ifdef CONFIG_NUMA p->mempolicy = mpol_copy(p->mempolicy); if (IS_ERR(p->mempolicy)) { diff -rup linux-2.6.7-rc3/kernel/sched.c linux-2.6.7-rc3-ss/kernel/sched.c --- linux-2.6.7-rc3/kernel/sched.c Mon Jun 7 12:14:56 2004 +++ linux-2.6.7-rc3-ss/kernel/sched.c Fri Jun 18 12:29:14 2004 @@ -40,6 +40,8 @@ #include #include #include +#include +#include #include @@ -217,6 +219,9 @@ struct runqueue { unsigned long expired_timestamp, nr_uninterruptible; unsigned long long timestamp_last_tick; task_t *curr, *idle; +#ifdef CONFIG_SCHEDSTATS + int cpu; /* to make easy reverse-lookups with per-cpu runqueues */ +#endif struct mm_struct *prev_mm; prio_array_t *active, *expired, arrays[2]; int best_expired_prio; @@ -232,6 +237,47 @@ struct runqueue { task_t *migration_thread; struct list_head migration_queue; #endif + +#ifdef CONFIG_SCHEDSTATS + /* latency stats */ + struct sched_info rq_sched_info; + + /* sys_sched_yield() stats */ + unsigned long yld_exp_empty; + unsigned long yld_act_empty; + unsigned long yld_both_empty; + unsigned long yld_cnt; + + /* schedule() stats */ + unsigned long sched_noswitch; + unsigned long sched_switch; + unsigned long sched_cnt; + + /* pull_task() stats */ + unsigned long pt_gained[MAX_IDLE_TYPES]; + unsigned long pt_lost[MAX_IDLE_TYPES]; + + /* active_load_balance() stats */ + unsigned long alb_cnt; + unsigned long alb_lost; + unsigned long alb_gained; + unsigned long alb_failed; + + /* try_to_wake_up() stats */ + unsigned long ttwu_cnt; + unsigned long ttwu_attempts; + unsigned long ttwu_moved; + + /* wake_up_forked_thread() stats */ + unsigned long wuft_cnt; + unsigned long wuft_moved; + + /* sched_migrate_task() stats */ + unsigned long smt_cnt; + + /* sched_balance_exec() stats */ + unsigned long sbe_cnt; +#endif }; static DEFINE_PER_CPU(struct runqueue, runqueues); @@ -278,6 +324,97 @@ static inline void task_rq_unlock(runque spin_unlock_irqrestore(&rq->lock, *flags); } +#ifdef CONFIG_SCHEDSTATS +/* + * 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 9 + +static int show_schedstat(struct seq_file *seq, void *v) +{ + int cpu; + enum idle_type itype; + + seq_printf(seq, "version %d\n", SCHEDSTAT_VERSION); + seq_printf(seq, "timestamp %lu\n", jiffies); + for_each_online_cpu (cpu) { + + int dcnt = 0; + + runqueue_t *rq = cpu_rq(cpu); + struct sched_domain *sd; + + /* runqueue-specific stats */ + seq_printf(seq, + "cpu%d %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu " + "%lu %lu %lu %lu %lu %lu %lu %lu %lu", + cpu, rq->yld_both_empty, + rq->yld_act_empty, rq->yld_exp_empty, + rq->yld_cnt, rq->sched_noswitch, + rq->sched_switch, rq->sched_cnt, + rq->alb_cnt, rq->alb_gained, rq->alb_lost, + rq->alb_failed, + rq->ttwu_cnt, rq->ttwu_moved, rq->ttwu_attempts, + rq->wuft_cnt, rq->wuft_moved, + rq->smt_cnt, rq->sbe_cnt, rq->rq_sched_info.cpu_time, + rq->rq_sched_info.run_delay, rq->rq_sched_info.pcnt); + + for (itype = IDLE; itype < MAX_IDLE_TYPES; itype++) + seq_printf(seq, " %lu %lu", rq->pt_gained[itype], + rq->pt_lost[itype]); + + seq_printf(seq, "\n"); + + /* domain-specific stats */ + for_each_domain(cpu, sd) { + char mask_str[NR_CPUS]; + + cpumask_scnprintf(mask_str, NR_CPUS, sd->span); + seq_printf(seq, "domain%d %s", dcnt++, mask_str); + for (itype = IDLE; itype < MAX_IDLE_TYPES; itype++) { + seq_printf(seq, " %lu %lu %lu %lu %lu", + sd->lb_cnt[itype], + sd->lb_failed[itype], + sd->lb_imbalance[itype], + sd->lb_nobusyq[itype], + sd->lb_nobusyg[itype]); + } + seq_printf(seq, " %lu %lu %lu %lu\n", + sd->sbe_pushed, sd->sbe_attempts, + sd->ttwu_wake_affine, sd->ttwu_wake_balance); + } + } + return 0; +} + +static int schedstat_open(struct inode *inode, struct file *file) +{ + unsigned size = PAGE_SIZE * (1 + num_online_cpus() / 32); + char *buf = kmalloc(size, GFP_KERNEL); + struct seq_file *m; + int res; + + if (!buf) + return -ENOMEM; + res = single_open(file, show_schedstat, NULL); + if (!res) { + m = file->private_data; + m->buf = buf; + m->size = size; + } else + kfree(buf); + return res; +} + +struct file_operations proc_schedstat_operations = { + .open = schedstat_open, + .read = seq_read, + .llseek = seq_lseek, + .release = single_release, +}; +#endif + /* * rq_lock - lock a given runqueue and disable interrupts. */ @@ -297,6 +434,113 @@ static inline void rq_unlock(runqueue_t spin_unlock_irq(&rq->lock); } +#ifdef CONFIG_SCHEDSTATS +/* + * Called when a process is dequeued from the active array and given + * the cpu. We should note that with the exception of interactive + * tasks, the expired queue will become the active queue after the active + * queue is empty, without explicitly dequeuing and requeuing tasks in the + * expired queue. (Interactive tasks may be requeued directly to the + * active queue, thus delaying tasks in the expired queue from running; + * see scheduler_tick()). + * + * This function is only called from sched_info_arrive(), rather than + * dequeue_task(). Even though a task may be queued and dequeued multiple + * times as it is shuffled about, we're really interested in knowing how + * long it was from the *first* time it was queued to the time that it + * finally hit a cpu. + */ +static inline void sched_info_dequeued(task_t *t) +{ + t->sched_info.last_queued = 0; +} + +/* + * Called when a task finally hits the cpu. We can now calculate how + * long it was waiting to run. We also note when it began so that we + * can keep stats on how long its timeslice is. + */ +static inline void sched_info_arrive(task_t *t) +{ + unsigned long now = jiffies; + unsigned long diff = 0; + struct runqueue *rq = task_rq(t); + + if (t->sched_info.last_queued) + diff = now - t->sched_info.last_queued; + sched_info_dequeued(t); + t->sched_info.run_delay += diff; + t->sched_info.last_arrival = now; + t->sched_info.pcnt++; + + if (!rq) + return; + + rq->rq_sched_info.run_delay += diff; + rq->rq_sched_info.pcnt++; +} + +/* + * Called when a process is queued into either the active or expired + * array. The time is noted and later used to determine how long we + * had to wait for us to reach the cpu. Since the expired queue will + * become the active queue after active queue is empty, without dequeuing + * and requeuing any tasks, we are interested in queuing to either. It + * is unusual but not impossible for tasks to be dequeued and immediately + * requeued in the same or another array: this can happen in sched_yield(), + * set_user_nice(), and even load_balance() as it moves tasks from runqueue + * to runqueue. + * + * This function is only called from enqueue_task(), but also only updates + * the timestamp if it is already not set. It's assumed that + * sched_info_dequeued() will clear that stamp when appropriate. + */ +static inline void sched_info_queued(task_t *t) +{ + if (!t->sched_info.last_queued) + t->sched_info.last_queued = jiffies; +} + +/* + * Called when a process ceases being the active-running process, either + * voluntarily or involuntarily. Now we can calculate how long we ran. + */ +static inline void sched_info_depart(task_t *t) +{ + struct runqueue *rq = task_rq(t); + unsigned long diff = jiffies - t->sched_info.last_arrival; + + t->sched_info.cpu_time += diff; + + if (rq) + rq->rq_sched_info.cpu_time += diff; +} + +/* + * Called when tasks are switched involuntarily due, typically, to expiring + * their time slice. (This may also be called when switching to or from + * the idle task.) We are only called when prev != next. + */ +static inline void sched_info_switch(task_t *prev, task_t *next) +{ + struct runqueue *rq = task_rq(prev); + + /* + * prev now departs the cpu. It's not interesting to record + * stats about how efficient we were at scheduling the idle + * process, however. + */ + if (prev != rq->idle) + sched_info_depart(prev); + + if (next != rq->idle) + sched_info_arrive(next); +} +#else +#define sched_info_queued(t) {} +#define sched_info_switch(t, next) {} +#endif /* CONFIG_SCHEDSTATS */ + /* * Adding/removing a task to/from a priority array: */ @@ -310,6 +554,7 @@ static void dequeue_task(struct task_str static void enqueue_task(struct task_struct *p, prio_array_t *array) { + sched_info_queued(p); list_add_tail(&p->run_list, array->queue + p->prio); __set_bit(p->prio, array->bitmap); array->nr_active++; @@ -735,11 +980,12 @@ static int try_to_wake_up(task_t * p, un runqueue_t *rq; #ifdef CONFIG_SMP unsigned long load, this_load; - struct sched_domain *sd; + struct sched_domain *sd = NULL; int new_cpu; #endif rq = task_rq_lock(p, &flags); + schedstat_inc(rq, ttwu_cnt); old_state = p->state; if (!(old_state & state)) goto out; @@ -788,23 +1034,35 @@ static int try_to_wake_up(task_t * p, un */ imbalance = sd->imbalance_pct + (sd->imbalance_pct - 100) / 2; - if ( ((sd->flags & SD_WAKE_AFFINE) && - !task_hot(p, rq->timestamp_last_tick, sd)) - || ((sd->flags & SD_WAKE_BALANCE) && - imbalance*this_load <= 100*load) ) { + if ((sd->flags & SD_WAKE_AFFINE) && + !task_hot(p, rq->timestamp_last_tick, sd)) { /* - * Now sd has SD_WAKE_AFFINE and p is cache cold in sd - * or sd has SD_WAKE_BALANCE and there is an imbalance + * This domain has SD_WAKE_AFFINE and p is cache cold + * in this domain. */ - if (cpu_isset(cpu, sd->span)) + if (cpu_isset(cpu, sd->span)) { + schedstat_inc(sd, ttwu_wake_affine); goto out_set_cpu; + } + } else if ((sd->flags & SD_WAKE_BALANCE) && + imbalance*this_load <= 100*load) { + /* + * This domain has SD_WAKE_BALANCE and there is + * an imbalance. + */ + if (cpu_isset(cpu, sd->span)) { + schedstat_inc(sd, ttwu_wake_balance); + goto out_set_cpu; + } } } new_cpu = cpu; /* Could not wake to this_cpu. Wake to cpu instead */ out_set_cpu: + schedstat_inc(rq, ttwu_attempts); new_cpu = wake_idle(new_cpu, p); if (new_cpu != cpu && cpu_isset(new_cpu, p->cpus_allowed)) { + schedstat_inc(rq, ttwu_moved); set_task_cpu(p, new_cpu); task_rq_unlock(rq, &flags); /* might preempt at this point */ @@ -856,7 +1114,7 @@ out: int fastcall wake_up_process(task_t * p) { return try_to_wake_up(p, TASK_STOPPED | - TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE, 0); + TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE, 0); } EXPORT_SYMBOL(wake_up_process); @@ -1156,13 +1414,6 @@ static void double_rq_unlock(runqueue_t spin_unlock(&rq2->lock); } -enum idle_type -{ - IDLE, - NOT_IDLE, - NEWLY_IDLE, -}; - #ifdef CONFIG_SMP /* @@ -1229,6 +1480,7 @@ void fastcall wake_up_forked_thread(task * Find the largest domain that this CPU is part of that * is willing to balance on clone: */ + schedstat_inc(this_rq, wuft_cnt); for_each_domain(this_cpu, tmp) if (tmp->flags & SD_BALANCE_CLONE) sd = tmp; @@ -1281,6 +1533,7 @@ lock_again: rq->nr_running++; } } else { + schedstat_inc(this_rq, wuft_moved); /* Not the local CPU - must adjust timestamp */ p->timestamp = (p->timestamp - this_rq->timestamp_last_tick) + rq->timestamp_last_tick; @@ -1311,6 +1564,7 @@ static void sched_migrate_task(task_t *p || unlikely(cpu_is_offline(dest_cpu))) goto out; + schedstat_inc(rq, smt_cnt); /* force the process onto the specified CPU */ if (migrate_task(p, dest_cpu, &req)) { /* Need to wait for migration thread (might exit: take ref). */ @@ -1338,6 +1592,7 @@ void sched_balance_exec(void) struct sched_domain *tmp, *sd = NULL; int new_cpu, this_cpu = get_cpu(); + schedstat_inc(this_rq(), sbe_cnt); /* Prefer the current CPU if there's only this task running */ if (this_rq()->nr_running <= 1) goto out; @@ -1346,9 +1601,11 @@ void sched_balance_exec(void) if (tmp->flags & SD_BALANCE_EXEC) sd = tmp; + schedstat_inc(sd, sbe_attempts); if (sd) { new_cpu = find_idlest_cpu(current, this_cpu, sd); if (new_cpu != this_cpu) { + schedstat_inc(sd, sbe_pushed); put_cpu(); sched_migrate_task(current, new_cpu); return; @@ -1487,6 +1744,15 @@ skip_queue: idx++; goto skip_bitmap; } + + /* + * Right now, this is the only place pull_task() is called, + * so we can safely collect pull_task() stats here rather than + * inside pull_task(). + */ + schedstat_inc(this_rq, pt_gained[idle]); + schedstat_inc(busiest, pt_lost[idle]); + pull_task(busiest, array, tmp, this_rq, dst_array, this_cpu); pulled++; @@ -1680,14 +1946,20 @@ static int load_balance(int this_cpu, ru int nr_moved; spin_lock(&this_rq->lock); - - group = find_busiest_group(sd, this_cpu, &imbalance, idle); - if (!group) - goto out_balanced; + schedstat_inc(sd, lb_cnt[idle]); + + group = find_busiest_group(sd, this_cpu, &imbalance, idle); + if (!group) { + schedstat_inc(sd, lb_nobusyg[idle]); + goto out_balanced; + } busiest = find_busiest_queue(group); - if (!busiest) - goto out_balanced; + if (!busiest) { + schedstat_inc(sd, lb_nobusyq[idle]); + goto out_balanced; + } + /* * This should be "impossible", but since load * balancing is inherently racy and statistical, @@ -1698,6 +1970,8 @@ static int load_balance(int this_cpu, ru goto out_balanced; } + schedstat_add(sd, lb_imbalance[idle], imbalance); + nr_moved = 0; if (busiest->nr_running > 1) { /* @@ -1714,6 +1988,7 @@ static int load_balance(int this_cpu, ru spin_unlock(&this_rq->lock); if (!nr_moved) { + schedstat_inc(sd, lb_failed[idle]); sd->nr_balance_failed++; if (unlikely(sd->nr_balance_failed > sd->cache_nice_tries+2)) { @@ -1768,17 +2043,23 @@ static int load_balance_newidle(int this unsigned long imbalance; int nr_moved = 0; + schedstat_inc(sd, lb_cnt[NEWLY_IDLE]); group = find_busiest_group(sd, this_cpu, &imbalance, NEWLY_IDLE); - if (!group) + if (!group) { + schedstat_inc(sd, lb_nobusyg[NEWLY_IDLE]); goto out; + } busiest = find_busiest_queue(group); - if (!busiest || busiest == this_rq) + if (!busiest || busiest == this_rq) { + schedstat_inc(sd, lb_nobusyq[NEWLY_IDLE]); goto out; + } /* Attempt to move tasks */ double_lock_balance(this_rq, busiest); + schedstat_add(sd, lb_imbalance[NEWLY_IDLE], imbalance); nr_moved = move_tasks(this_rq, this_cpu, busiest, imbalance, sd, NEWLY_IDLE); @@ -1820,6 +2101,7 @@ static void active_load_balance(runqueue struct sched_group *group, *busy_group; int i; + schedstat_inc(busiest, alb_cnt); if (busiest->nr_running <= 1) return; @@ -1866,7 +2148,12 @@ static void active_load_balance(runqueue if (unlikely(busiest == rq)) goto next_group; double_lock_balance(busiest, rq); - move_tasks(rq, push_cpu, busiest, 1, sd, IDLE); + if (move_tasks(rq, push_cpu, busiest, 1, sd, IDLE)) { + schedstat_inc(busiest, alb_lost); + schedstat_inc(rq, alb_gained); + } else { + schedstat_inc(busiest, alb_failed); + } spin_unlock(&rq->lock); next_group: group = group->next; @@ -2210,6 +2497,7 @@ need_resched: rq = this_rq(); release_kernel_lock(prev); + schedstat_inc(rq, sched_cnt); now = sched_clock(); if (likely(now - prev->timestamp < NS_MAX_SLEEP_AVG)) run_time = now - prev->timestamp; @@ -2256,12 +2544,14 @@ need_resched: /* * Switch the active and expired arrays. */ + schedstat_inc(rq, sched_switch); rq->active = rq->expired; rq->expired = array; array = rq->active; rq->expired_timestamp = 0; rq->best_expired_prio = MAX_PRIO; - } + } else + schedstat_inc(rq, sched_noswitch); idx = sched_find_first_bit(array->bitmap); queue = array->queue + idx; @@ -2297,6 +2587,7 @@ switch_tasks: } prev->timestamp = now; + sched_info_switch(prev, next); if (likely(prev != next)) { next->timestamp = now; rq->nr_switches++; @@ -2993,6 +3284,7 @@ asmlinkage long sys_sched_yield(void) prio_array_t *array = current->array; prio_array_t *target = rq->expired; + schedstat_inc(rq, yld_cnt); /* * We implement yielding by moving the task into the expired * queue. @@ -3003,6 +3295,15 @@ asmlinkage long sys_sched_yield(void) if (unlikely(rt_task(current))) target = rq->active; + if (current->array->nr_active == 1) { + schedstat_inc(rq, yld_act_empty); + if (!rq->expired->nr_active) { + schedstat_inc(rq, yld_both_empty); + } + } else if (!rq->expired->nr_active) { + schedstat_inc(rq, yld_exp_empty); + } + dequeue_task(current, array); enqueue_task(current, target); @@ -3587,7 +3888,7 @@ static int migration_call(struct notifie rq->idle->static_prio = MAX_PRIO; __setscheduler(rq->idle, SCHED_NORMAL, 0); task_rq_unlock(rq, &flags); - BUG_ON(rq->nr_running != 0); + BUG_ON(rq->nr_running != 0); /* No need to migrate the tasks: it was best-effort if * they didn't do lock_cpu_hotplug(). Just wake up @@ -3602,7 +3903,7 @@ static int migration_call(struct notifie complete(&req->done); } spin_unlock_irq(&rq->lock); - break; + break; #endif } return NOTIFY_OK; @@ -3923,6 +4224,9 @@ void __init sched_init(void) spin_lock_init(&rq->lock); rq->active = rq->arrays; rq->expired = rq->arrays + 1; +#ifdef CONFIG_SCHEDSTATS + rq->cpu = i; +#endif /* CONFIG_SCHEDSTATS */ rq->best_expired_prio = MAX_PRIO; #ifdef CONFIG_SMP