From: Martin Schwidefsky <[email protected]>
The cpu idle field in the output of /proc/stat is too small for cpus
that have been idle for more than a tick. Add the architecture hook
arch_idle_time that allows to add the not accounted idle time of a
sleeping cpu without waking the cpu.
The s390 implementation of arch_idle_time uses the already existing
s390_idle_data per_cpu variable to find the sleep time of a neighboring
idle cpu.
Signed-off-by: Martin Schwidefsky <[email protected]>
---
arch/s390/include/asm/cputime.h | 4 ++++
arch/s390/kernel/vtime.c | 16 ++++++++++++++++
fs/proc/stat.c | 8 ++++++--
3 files changed, 26 insertions(+), 2 deletions(-)
Index: linux-2.6/arch/s390/include/asm/cputime.h
===================================================================
--- linux-2.6.orig/arch/s390/include/asm/cputime.h
+++ linux-2.6/arch/s390/include/asm/cputime.h
@@ -174,4 +174,8 @@ cputime64_to_clock_t(cputime64_t cputime
return __div(cputime, 4096000000ULL / USER_HZ);
}
+cputime64_t s390_get_idle_time(int cpu);
+
+#define arch_idle_time(cpu) s390_get_idle_time(cpu)
+
#endif /* _S390_CPUTIME_H */
Index: linux-2.6/arch/s390/kernel/vtime.c
===================================================================
--- linux-2.6.orig/arch/s390/kernel/vtime.c
+++ linux-2.6/arch/s390/kernel/vtime.c
@@ -238,6 +238,22 @@ void vtime_stop_cpu(void)
}
}
+cputime64_t s390_get_idle_time(int cpu)
+{
+ struct s390_idle_data *idle;
+ unsigned long long now, idle_time, idle_enter;
+
+ idle = &per_cpu(s390_idle, cpu);
+ spin_lock(&idle->lock);
+ now = get_clock();
+ idle_time = 0;
+ idle_enter = idle->idle_enter;
+ if (idle_enter != 0ULL && idle_enter < now)
+ idle_time = now - idle_enter;
+ spin_unlock(&idle->lock);
+ return idle_time;
+}
+
/*
* Sorted add to a list. List is linear searched until first bigger
* element is found.
Index: linux-2.6/fs/proc/stat.c
===================================================================
--- linux-2.6.orig/fs/proc/stat.c
+++ linux-2.6/fs/proc/stat.c
@@ -18,6 +18,9 @@
#ifndef arch_irq_stat
#define arch_irq_stat() 0
#endif
+#ifndef arch_idle_time
+#define arch_idle_time(cpu) 0
+#endif
static int show_stat(struct seq_file *p, void *v)
{
@@ -39,7 +42,8 @@ static int show_stat(struct seq_file *p,
user = cputime64_add(user, kstat_cpu(i).cpustat.user);
nice = cputime64_add(nice, kstat_cpu(i).cpustat.nice);
system = cputime64_add(system, kstat_cpu(i).cpustat.system);
- idle = cputime64_add(idle, kstat_cpu(i).cpustat.idle);
+ idle = cputime64_add(idle, kstat_cpu(i).cpustat.idle +
+ arch_idle_time(i));
iowait = cputime64_add(iowait, kstat_cpu(i).cpustat.iowait);
irq = cputime64_add(irq, kstat_cpu(i).cpustat.irq);
softirq = cputime64_add(softirq, kstat_cpu(i).cpustat.softirq);
@@ -68,7 +72,7 @@ static int show_stat(struct seq_file *p,
user = kstat_cpu(i).cpustat.user;
nice = kstat_cpu(i).cpustat.nice;
system = kstat_cpu(i).cpustat.system;
- idle = kstat_cpu(i).cpustat.idle;
+ idle = kstat_cpu(i).cpustat.idle + arch_idle_time(i);
iowait = kstat_cpu(i).cpustat.iowait;
irq = kstat_cpu(i).cpustat.irq;
softirq = kstat_cpu(i).cpustat.softirq;
On Tue, 14 Apr 2009 14:18:51 +0200
Martin Schwidefsky <[email protected]> wrote:
> --- linux-2.6.orig/fs/proc/stat.c
> +++ linux-2.6/fs/proc/stat.c
> @@ -18,6 +18,9 @@
> #ifndef arch_irq_stat
> #define arch_irq_stat() 0
> #endif
> +#ifndef arch_idle_time
> +#define arch_idle_time(cpu) 0
> +#endif
>
> static int show_stat(struct seq_file *p, void *v)
> {
> @@ -39,7 +42,8 @@ static int show_stat(struct seq_file *p,
> user = cputime64_add(user, kstat_cpu(i).cpustat.user);
> nice = cputime64_add(nice, kstat_cpu(i).cpustat.nice);
> system = cputime64_add(system, kstat_cpu(i).cpustat.system);
> - idle = cputime64_add(idle, kstat_cpu(i).cpustat.idle);
> + idle = cputime64_add(idle, kstat_cpu(i).cpustat.idle +
> + arch_idle_time(i));
> iowait = cputime64_add(iowait, kstat_cpu(i).cpustat.iowait);
> irq = cputime64_add(irq, kstat_cpu(i).cpustat.irq);
> softirq = cputime64_add(softirq, kstat_cpu(i).cpustat.softirq);
> @@ -68,7 +72,7 @@ static int show_stat(struct seq_file *p,
> user = kstat_cpu(i).cpustat.user;
> nice = kstat_cpu(i).cpustat.nice;
> system = kstat_cpu(i).cpustat.system;
> - idle = kstat_cpu(i).cpustat.idle;
> + idle = kstat_cpu(i).cpustat.idle + arch_idle_time(i);
> iowait = kstat_cpu(i).cpustat.iowait;
> irq = kstat_cpu(i).cpustat.irq;
> softirq = kstat_cpu(i).cpustat.softirq;
Looks OK to me. Please merge it via the s390 tree.
Do other architectures need to fix this?
On Fri, 17 Apr 2009 16:02:11 -0700
Andrew Morton <[email protected]> wrote:
> Looks OK to me. Please merge it via the s390 tree.
>
> Do other architectures need to fix this?
Thanks Andrew for the quick review. Other architecture might want to
implement this as well, with tick-less the idle numbers in /proc/stat
are just way off. The implementation of the arch_idle_time is not easy
I'm afraid, the sleep time of each cpu needs to be measured in a way
that it can be read from neighbouring cpus (and it should better be a
precise). Not something I would want to have to implement for x86.
--
blue skies,
Martin.
"Reality continues to ruin my life." - Calvin.