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: <asm/paravirt.h> doesn't exist on many archs]

Signed-off-by: Glauber Costa <glommer@redhat.com>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Tested-by: Eric B Munson <emunson@mgebm.net>
CC: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
CC: Anthony Liguori <aliguori@us.ibm.com>
Signed-off-by: Avi Kivity <avi@redhat.com>

diff --git a/kernel/sched.c b/kernel/sched.c
index 3f2e502d609bb7957e10b4ec6aa647c9b3069c6e..f98a28b19b2aee75716130dd0457e69e26932ce4 100644 (file)
--- a/kernel/sched.c
+++ b/kernel/sched.c
@@ -75,6 +75,9 @@
 #include <asm/tlb.h>
 #include <asm/irq_regs.h>
 #include <asm/mutex.h>
+#ifdef CONFIG_PARAVIRT
+#include <asm/paravirt.h>
+#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(&paravirt_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))