From e6e6685accfa81f509fadfc9624bc7c3862d75c4 Mon Sep 17 00:00:00 2001 From: Glauber Costa Date: Mon, 11 Jul 2011 15:28:17 -0400 Subject: KVM guest: Steal time accounting This patch accounts steal time time in account_process_tick. If one or more tick is considered stolen in the current accounting cycle, user/system accounting is skipped. Idle is fine, since the hypervisor does not report steal time if the guest is halted. Accounting steal time from the core scheduler give us the advantage of direct acess to the runqueue data. In a later opportunity, it can be used to tweak cpu power and make the scheduler aware of the time it lost. [avi: doesn't exist on many archs] Signed-off-by: Glauber Costa Acked-by: Rik van Riel Acked-by: Peter Zijlstra Tested-by: Eric B Munson CC: Jeremy Fitzhardinge CC: Anthony Liguori Signed-off-by: Avi Kivity --- kernel/sched.c | 43 +++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 43 insertions(+) (limited to 'kernel/sched.c') diff --git a/kernel/sched.c b/kernel/sched.c index 3f2e502d609b..f98a28b19b2a 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -75,6 +75,9 @@ #include #include #include +#ifdef CONFIG_PARAVIRT +#include +#endif #include "sched_cpupri.h" #include "workqueue_sched.h" @@ -528,6 +531,9 @@ struct rq { #ifdef CONFIG_IRQ_TIME_ACCOUNTING u64 prev_irq_time; #endif +#ifdef CONFIG_PARAVIRT + u64 prev_steal_time; +#endif /* calc_load related fields */ unsigned long calc_load_update; @@ -1953,6 +1959,18 @@ void account_system_vtime(struct task_struct *curr) } EXPORT_SYMBOL_GPL(account_system_vtime); +#endif /* CONFIG_IRQ_TIME_ACCOUNTING */ + +#ifdef CONFIG_PARAVIRT +static inline u64 steal_ticks(u64 steal) +{ + if (unlikely(steal > NSEC_PER_SEC)) + return div_u64(steal, TICK_NSEC); + + return __iter_div_u64_rem(steal, TICK_NSEC, &steal); +} +#endif + static void update_rq_clock_task(struct rq *rq, s64 delta) { s64 irq_delta; @@ -3845,6 +3863,25 @@ void account_idle_time(cputime_t cputime) cpustat->idle = cputime64_add(cpustat->idle, cputime64); } +static __always_inline bool steal_account_process_tick(void) +{ +#ifdef CONFIG_PARAVIRT + if (static_branch(¶virt_steal_enabled)) { + u64 steal, st = 0; + + steal = paravirt_steal_clock(smp_processor_id()); + steal -= this_rq()->prev_steal_time; + + st = steal_ticks(steal); + this_rq()->prev_steal_time += st * TICK_NSEC; + + account_steal_time(st); + return st; + } +#endif + return false; +} + #ifndef CONFIG_VIRT_CPU_ACCOUNTING #ifdef CONFIG_IRQ_TIME_ACCOUNTING @@ -3876,6 +3913,9 @@ static void irqtime_account_process_tick(struct task_struct *p, int user_tick, cputime64_t tmp = cputime_to_cputime64(cputime_one_jiffy); struct cpu_usage_stat *cpustat = &kstat_this_cpu.cpustat; + if (steal_account_process_tick()) + return; + if (irqtime_account_hi_update()) { cpustat->irq = cputime64_add(cpustat->irq, tmp); } else if (irqtime_account_si_update()) { @@ -3929,6 +3969,9 @@ void account_process_tick(struct task_struct *p, int user_tick) return; } + if (steal_account_process_tick()) + return; + if (user_tick) account_user_time(p, cputime_one_jiffy, one_jiffy_scaled); else if ((p != rq->idle) || (irq_count() != HARDIRQ_OFFSET)) -- cgit v1.2.3 From 095c0aa83e52d6c3dd7168610746703921f570af Mon Sep 17 00:00:00 2001 From: Glauber Costa Date: Mon, 11 Jul 2011 15:28:18 -0400 Subject: sched: adjust scheduler cpu power for stolen time This patch makes update_rq_clock() aware of steal time. The mechanism of operation is not different from irq_time, and follows the same principles. This lives in a CONFIG option itself, and can be compiled out independently of the rest of steal time reporting. The effect of disabling it is that the scheduler will still report steal time (that cannot be disabled), but won't use this information for cpu power adjustments. Everytime update_rq_clock_task() is invoked, we query information about how much time was stolen since last call, and feed it into sched_rt_avg_update(). Although steal time reporting in account_process_tick() keeps track of the last time we read the steal clock, in prev_steal_time, this patch do it independently using another field, prev_steal_time_rq. This is because otherwise, information about time accounted in update_process_tick() would never reach us in update_rq_clock(). Signed-off-by: Glauber Costa Acked-by: Rik van Riel Acked-by: Peter Zijlstra Tested-by: Eric B Munson CC: Jeremy Fitzhardinge CC: Anthony Liguori Signed-off-by: Avi Kivity --- arch/x86/Kconfig | 12 ++++++++++++ kernel/sched.c | 47 +++++++++++++++++++++++++++++++++++++---------- kernel/sched_features.h | 4 ++-- 3 files changed, 51 insertions(+), 12 deletions(-) (limited to 'kernel/sched.c') diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig index da349723d411..1f03e221a01e 100644 --- a/arch/x86/Kconfig +++ b/arch/x86/Kconfig @@ -512,6 +512,18 @@ menuconfig PARAVIRT_GUEST if PARAVIRT_GUEST +config PARAVIRT_TIME_ACCOUNTING + bool "Paravirtual steal time accounting" + select PARAVIRT + default n + ---help--- + Select this option to enable fine granularity task steal time + accounting. Time spent executing other tasks in parallel with + the current vCPU is discounted from the vCPU power. To account for + that, there can be a small performance impact. + + If in doubt, say N here. + source "arch/x86/xen/Kconfig" config KVM_CLOCK diff --git a/kernel/sched.c b/kernel/sched.c index f98a28b19b2a..b35ac50b26c8 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -534,6 +534,9 @@ struct rq { #ifdef CONFIG_PARAVIRT u64 prev_steal_time; #endif +#ifdef CONFIG_PARAVIRT_TIME_ACCOUNTING + u64 prev_steal_time_rq; +#endif /* calc_load related fields */ unsigned long calc_load_update; @@ -1973,8 +1976,14 @@ static inline u64 steal_ticks(u64 steal) static void update_rq_clock_task(struct rq *rq, s64 delta) { - s64 irq_delta; - +/* + * In theory, the compile should just see 0 here, and optimize out the call + * to sched_rt_avg_update. But I don't trust it... + */ +#if defined(CONFIG_IRQ_TIME_ACCOUNTING) || defined(CONFIG_PARAVIRT_TIME_ACCOUNTING) + s64 steal = 0, irq_delta = 0; +#endif +#ifdef CONFIG_IRQ_TIME_ACCOUNTING irq_delta = irq_time_read(cpu_of(rq)) - rq->prev_irq_time; /* @@ -1997,12 +2006,35 @@ static void update_rq_clock_task(struct rq *rq, s64 delta) rq->prev_irq_time += irq_delta; delta -= irq_delta; +#endif +#ifdef CONFIG_PARAVIRT_TIME_ACCOUNTING + if (static_branch((¶virt_steal_rq_enabled))) { + u64 st; + + steal = paravirt_steal_clock(cpu_of(rq)); + steal -= rq->prev_steal_time_rq; + + if (unlikely(steal > delta)) + steal = delta; + + st = steal_ticks(steal); + steal = st * TICK_NSEC; + + rq->prev_steal_time_rq += steal; + + delta -= steal; + } +#endif + rq->clock_task += delta; - if (irq_delta && sched_feat(NONIRQ_POWER)) - sched_rt_avg_update(rq, irq_delta); +#if defined(CONFIG_IRQ_TIME_ACCOUNTING) || defined(CONFIG_PARAVIRT_TIME_ACCOUNTING) + if ((irq_delta + steal) && sched_feat(NONTASK_POWER)) + sched_rt_avg_update(rq, irq_delta + steal); +#endif } +#ifdef CONFIG_IRQ_TIME_ACCOUNTING static int irqtime_account_hi_update(void) { struct cpu_usage_stat *cpustat = &kstat_this_cpu.cpustat; @@ -2037,12 +2069,7 @@ static int irqtime_account_si_update(void) #define sched_clock_irqtime (0) -static void update_rq_clock_task(struct rq *rq, s64 delta) -{ - rq->clock_task += delta; -} - -#endif /* CONFIG_IRQ_TIME_ACCOUNTING */ +#endif #include "sched_idletask.c" #include "sched_fair.c" diff --git a/kernel/sched_features.h b/kernel/sched_features.h index be40f7371ee1..ca3b025f8669 100644 --- a/kernel/sched_features.h +++ b/kernel/sched_features.h @@ -61,9 +61,9 @@ SCHED_FEAT(LB_BIAS, 1) SCHED_FEAT(OWNER_SPIN, 1) /* - * Decrement CPU power based on irq activity + * Decrement CPU power based on time not spent running tasks */ -SCHED_FEAT(NONIRQ_POWER, 1) +SCHED_FEAT(NONTASK_POWER, 1) /* * Queue remote wakeups on the target CPU and process them -- cgit v1.2.3