From aa9c4c0f967fdb482ea95e8473ec3d201e6e0781 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Wed, 17 Dec 2008 14:10:57 +0100 Subject: perfcounters: fix task clock counter Impact: fix per task clock counter precision Signed-off-by: Ingo Molnar --- include/linux/kernel_stat.h | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'include/linux/kernel_stat.h') diff --git a/include/linux/kernel_stat.h b/include/linux/kernel_stat.h index 4a145caeee0..1b2e3242497 100644 --- a/include/linux/kernel_stat.h +++ b/include/linux/kernel_stat.h @@ -66,7 +66,15 @@ static inline unsigned int kstat_irqs(unsigned int irq) return sum; } + +/* + * Lock/unlock the current runqueue - to extract task statistics: + */ +extern void curr_rq_lock_irq_save(unsigned long *flags); +extern void curr_rq_unlock_irq_restore(unsigned long *flags); +extern unsigned long long __task_delta_exec(struct task_struct *tsk, int update); extern unsigned long long task_delta_exec(struct task_struct *); + extern void account_user_time(struct task_struct *, cputime_t); extern void account_user_time_scaled(struct task_struct *, cputime_t); extern void account_system_time(struct task_struct *, int, cputime_t); -- cgit v1.2.3-70-g09d2 From 849691a6cd40270ff5f4a8846d5f6bf8df663ffc Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Mon, 6 Apr 2009 11:45:12 +0200 Subject: perf_counter: remove rq->lock usage Now that all the task runtime clock users are gone, remove the ugly rq->lock usage from perf counters, which solves the nasty deadlock seen when a software task clock counter was read from an NMI overflow context. Signed-off-by: Peter Zijlstra Cc: Paul Mackerras Cc: Corey Ashford LKML-Reference: <20090406094518.531137582@chello.nl> Signed-off-by: Ingo Molnar --- include/linux/kernel_stat.h | 2 -- kernel/perf_counter.c | 42 ++++++++++++++++-------------------------- kernel/sched.c | 20 -------------------- 3 files changed, 16 insertions(+), 48 deletions(-) (limited to 'include/linux/kernel_stat.h') diff --git a/include/linux/kernel_stat.h b/include/linux/kernel_stat.h index b6d2887a5d8..080d1fd461d 100644 --- a/include/linux/kernel_stat.h +++ b/include/linux/kernel_stat.h @@ -85,8 +85,6 @@ static inline unsigned int kstat_irqs(unsigned int irq) /* * Lock/unlock the current runqueue - to extract task statistics: */ -extern void curr_rq_lock_irq_save(unsigned long *flags); -extern void curr_rq_unlock_irq_restore(unsigned long *flags); extern unsigned long long __task_delta_exec(struct task_struct *tsk, int update); extern unsigned long long task_delta_exec(struct task_struct *); diff --git a/kernel/perf_counter.c b/kernel/perf_counter.c index 56b7eb53d67..f4f7596f784 100644 --- a/kernel/perf_counter.c +++ b/kernel/perf_counter.c @@ -172,8 +172,7 @@ static void __perf_counter_remove_from_context(void *info) if (ctx->task && cpuctx->task_ctx != ctx) return; - curr_rq_lock_irq_save(&flags); - spin_lock(&ctx->lock); + spin_lock_irqsave(&ctx->lock, flags); counter_sched_out(counter, cpuctx, ctx); @@ -198,8 +197,7 @@ static void __perf_counter_remove_from_context(void *info) perf_max_counters - perf_reserved_percpu); } - spin_unlock(&ctx->lock); - curr_rq_unlock_irq_restore(&flags); + spin_unlock_irqrestore(&ctx->lock, flags); } @@ -319,8 +317,7 @@ static void __perf_counter_disable(void *info) if (ctx->task && cpuctx->task_ctx != ctx) return; - curr_rq_lock_irq_save(&flags); - spin_lock(&ctx->lock); + spin_lock_irqsave(&ctx->lock, flags); /* * If the counter is on, turn it off. @@ -336,8 +333,7 @@ static void __perf_counter_disable(void *info) counter->state = PERF_COUNTER_STATE_OFF; } - spin_unlock(&ctx->lock); - curr_rq_unlock_irq_restore(&flags); + spin_unlock_irqrestore(&ctx->lock, flags); } /* @@ -515,8 +511,7 @@ static void __perf_install_in_context(void *info) if (ctx->task && cpuctx->task_ctx != ctx) return; - curr_rq_lock_irq_save(&flags); - spin_lock(&ctx->lock); + spin_lock_irqsave(&ctx->lock, flags); update_context_time(ctx); /* @@ -565,8 +560,7 @@ static void __perf_install_in_context(void *info) unlock: hw_perf_restore(perf_flags); - spin_unlock(&ctx->lock); - curr_rq_unlock_irq_restore(&flags); + spin_unlock_irqrestore(&ctx->lock, flags); } /* @@ -641,8 +635,7 @@ static void __perf_counter_enable(void *info) if (ctx->task && cpuctx->task_ctx != ctx) return; - curr_rq_lock_irq_save(&flags); - spin_lock(&ctx->lock); + spin_lock_irqsave(&ctx->lock, flags); update_context_time(ctx); counter->prev_state = counter->state; @@ -678,8 +671,7 @@ static void __perf_counter_enable(void *info) } unlock: - spin_unlock(&ctx->lock); - curr_rq_unlock_irq_restore(&flags); + spin_unlock_irqrestore(&ctx->lock, flags); } /* @@ -971,7 +963,7 @@ int perf_counter_task_disable(void) if (likely(!ctx->nr_counters)) return 0; - curr_rq_lock_irq_save(&flags); + local_irq_save(flags); cpu = smp_processor_id(); perf_counter_task_sched_out(curr, cpu); @@ -992,9 +984,7 @@ int perf_counter_task_disable(void) hw_perf_restore(perf_flags); - spin_unlock(&ctx->lock); - - curr_rq_unlock_irq_restore(&flags); + spin_unlock_irqrestore(&ctx->lock, flags); return 0; } @@ -1011,7 +1001,7 @@ int perf_counter_task_enable(void) if (likely(!ctx->nr_counters)) return 0; - curr_rq_lock_irq_save(&flags); + local_irq_save(flags); cpu = smp_processor_id(); perf_counter_task_sched_out(curr, cpu); @@ -1037,7 +1027,7 @@ int perf_counter_task_enable(void) perf_counter_task_sched_in(curr, cpu); - curr_rq_unlock_irq_restore(&flags); + local_irq_restore(flags); return 0; } @@ -1095,12 +1085,12 @@ static void __read(void *info) struct perf_counter_context *ctx = counter->ctx; unsigned long flags; - curr_rq_lock_irq_save(&flags); + local_irq_save(flags); if (ctx->is_active) update_context_time(ctx); counter->hw_ops->read(counter); update_counter_times(counter); - curr_rq_unlock_irq_restore(&flags); + local_irq_restore(flags); } static u64 perf_counter_read(struct perf_counter *counter) @@ -2890,7 +2880,7 @@ __perf_counter_exit_task(struct task_struct *child, * Be careful about zapping the list - IRQ/NMI context * could still be processing it: */ - curr_rq_lock_irq_save(&flags); + local_irq_save(flags); perf_flags = hw_perf_save_disable(); cpuctx = &__get_cpu_var(perf_cpu_context); @@ -2903,7 +2893,7 @@ __perf_counter_exit_task(struct task_struct *child, child_ctx->nr_counters--; hw_perf_restore(perf_flags); - curr_rq_unlock_irq_restore(&flags); + local_irq_restore(flags); } parent_counter = child_counter->parent; diff --git a/kernel/sched.c b/kernel/sched.c index f76e3c0188a..0de2f814fb1 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -997,26 +997,6 @@ static struct rq *task_rq_lock(struct task_struct *p, unsigned long *flags) } } -void curr_rq_lock_irq_save(unsigned long *flags) - __acquires(rq->lock) -{ - struct rq *rq; - - local_irq_save(*flags); - rq = cpu_rq(smp_processor_id()); - spin_lock(&rq->lock); -} - -void curr_rq_unlock_irq_restore(unsigned long *flags) - __releases(rq->lock) -{ - struct rq *rq; - - rq = cpu_rq(smp_processor_id()); - spin_unlock(&rq->lock); - local_irq_restore(*flags); -} - void task_rq_unlock_wait(struct task_struct *p) { struct rq *rq = task_rq(p); -- cgit v1.2.3-70-g09d2 From f1f9b3b1795da8625e0e6096813c9d18d4a344ce Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Mon, 20 Apr 2009 20:38:21 +0200 Subject: perfcounters, sched: remove __task_delta_exec() This function was left orphan by the latest round of sw-counter cleanups. [ Impact: remove unused kernel function ] Signed-off-by: Ingo Molnar --- include/linux/kernel_stat.h | 1 - kernel/sched.c | 23 ----------------------- 2 files changed, 24 deletions(-) (limited to 'include/linux/kernel_stat.h') diff --git a/include/linux/kernel_stat.h b/include/linux/kernel_stat.h index 080d1fd461d..a77c6007dc9 100644 --- a/include/linux/kernel_stat.h +++ b/include/linux/kernel_stat.h @@ -85,7 +85,6 @@ static inline unsigned int kstat_irqs(unsigned int irq) /* * Lock/unlock the current runqueue - to extract task statistics: */ -extern unsigned long long __task_delta_exec(struct task_struct *tsk, int update); extern unsigned long long task_delta_exec(struct task_struct *); extern void account_user_time(struct task_struct *, cputime_t, cputime_t); diff --git a/kernel/sched.c b/kernel/sched.c index b66a08c2480..a69278eef42 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -4546,29 +4546,6 @@ DEFINE_PER_CPU(struct kernel_stat, kstat); EXPORT_PER_CPU_SYMBOL(kstat); -/* - * Return any ns on the sched_clock that have not yet been banked in - * @p in case that task is currently running. - */ -unsigned long long __task_delta_exec(struct task_struct *p, int update) -{ - s64 delta_exec; - struct rq *rq; - - rq = task_rq(p); - WARN_ON_ONCE(!runqueue_is_locked()); - WARN_ON_ONCE(!task_current(rq, p)); - - if (update) - update_rq_clock(rq); - - delta_exec = rq->clock - p->se.exec_start; - - WARN_ON_ONCE(delta_exec < 0); - - return delta_exec; -} - /* * Return any ns on the sched_clock that have not yet been banked in * @p in case that task is currently running. -- cgit v1.2.3-70-g09d2 From aa0ce5bbc2dbb1853bd0c6d13f17716fcc38ac5a Mon Sep 17 00:00:00 2001 From: Keika Kobayashi Date: Wed, 17 Jun 2009 16:25:52 -0700 Subject: softirq: introduce statistics for softirq Statistics for softirq doesn't exist. It will be helpful like statistics for interrupts. This patch introduces counting the number of softirq, which will be exported in /proc/softirqs. When softirq handler consumes much CPU time, /proc/stat is like the following. $ while :; do cat /proc/stat | head -n1 ; sleep 10 ; done cpu 88 0 408 739665 583 28 2 0 0 cpu 450 0 1090 740970 594 28 1294 0 0 ^^^^ softirq In such a situation, /proc/softirqs shows us which softirq handler is invoked. We can see the increase rate of softirqs. $ cat /proc/softirqs CPU0 CPU1 CPU2 CPU3 HI 0 0 0 0 TIMER 462850 462805 462782 462718 NET_TX 0 0 0 365 NET_RX 2472 2 2 40 BLOCK 0 0 381 1164 TASKLET 0 0 0 224 SCHED 462654 462689 462698 462427 RCU 3046 2423 3367 3173 $ cat /proc/softirqs CPU0 CPU1 CPU2 CPU3 HI 0 0 0 0 TIMER 463361 465077 465056 464991 NET_TX 53 0 1 365 NET_RX 3757 2 2 40 BLOCK 0 0 398 1170 TASKLET 0 0 0 224 SCHED 463074 464318 464612 463330 RCU 3505 2948 3947 3673 When CPU TIME of softirq is high, the rates of increase is the following. TIMER : 220/sec : CPU1-3 NET_TX : 5/sec : CPU0 NET_RX : 120/sec : CPU0 SCHED : 40-200/sec : all CPU RCU : 45-58/sec : all CPU The rates of increase in an idle mode is the following. TIMER : 250/sec SCHED : 250/sec RCU : 2/sec It seems many softirqs for receiving packets and rcu are invoked. This gives us help for checking system. Signed-off-by: Keika Kobayashi Reviewed-by: Hiroshi Shimamoto Reviewed-by: KOSAKI Motohiro Cc: Ingo Molnar Cc: Eric Dumazet Cc: Alexey Dobriyan Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- include/linux/kernel_stat.h | 12 ++++++++++++ kernel/softirq.c | 1 + 2 files changed, 13 insertions(+) (limited to 'include/linux/kernel_stat.h') diff --git a/include/linux/kernel_stat.h b/include/linux/kernel_stat.h index a77c6007dc9..348fa8874b5 100644 --- a/include/linux/kernel_stat.h +++ b/include/linux/kernel_stat.h @@ -5,6 +5,7 @@ #include #include #include +#include #include #include @@ -31,6 +32,7 @@ struct kernel_stat { #ifndef CONFIG_GENERIC_HARDIRQS unsigned int irqs[NR_IRQS]; #endif + unsigned int softirqs[NR_SOFTIRQS]; }; DECLARE_PER_CPU(struct kernel_stat, kstat); @@ -67,6 +69,16 @@ extern unsigned int kstat_irqs_cpu(unsigned int irq, int cpu); #endif +static inline void kstat_incr_softirqs_this_cpu(unsigned int irq) +{ + kstat_this_cpu.softirqs[irq]++; +} + +static inline unsigned int kstat_softirqs_cpu(unsigned int irq, int cpu) +{ + return kstat_cpu(cpu).softirqs[irq]; +} + /* * Number of interrupts per specific IRQ source, since bootup */ diff --git a/kernel/softirq.c b/kernel/softirq.c index b41fb710e11..3a94905fa5d 100644 --- a/kernel/softirq.c +++ b/kernel/softirq.c @@ -213,6 +213,7 @@ restart: do { if (pending & 1) { int prev_count = preempt_count(); + kstat_incr_softirqs_this_cpu(h - softirq_vec); trace_softirq_entry(h, softirq_vec); h->action(h); -- cgit v1.2.3-70-g09d2