Subject: Provide individual CPU usage measurement based on idle time
From: Carsten Emde <C.Emde@osadl.org>
Date: Sun, 22 Apr 2012 15:19:46 +0100

The various methods to determine CPU usage and load have a number of
disadvantages (see also Documentation/cpu-load.txt), and a
straight-forward method to gain usage information about a particular CPU
is lacking. However, in the context of setting CPU affinity and
isolation, it is often required to monitor the effective usage ratio of
a CPU.

This patch adds an additional CPU usage measuring method that is based
on idle time processing. The data are available for every CPU in
/proc/cpuload/cpuN/data in the format "<idletime> <runtime>". The
counters can be reset by writing to /proc/cpuload/resetall for all CPUs
and to /proc/cpuload/cpuN/reset for a particular CPU, respectively.

To calculate the CPU usage since the most recent reset, divide the
runtime by the sum of runtime plus idletime, e.g.

# for i in `ls -1d /proc/cpuload/cpu* | sort -nk1.18`
> do
>   echo "$i: `awk '{ print (100.0*$2) / ($1+$2)"%" }' <$i/data`"
>   echo 1 >$i/reset
> done
/proc/cpuload/cpu0: 72.0048%
/proc/cpuload/cpu1: 5.49522%
/proc/cpuload/cpu2: 0.27916%
/proc/cpuload/cpu3: 32.3493%

Signed-off-by: Carsten Emde <C.Emde@osadl.org>

---
 init/Kconfig                   |   12 +++
 kernel/sched/Makefile          |    1 
 kernel/sched/core.c            |   31 +++++++++
 kernel/sched/cpu_idleruntime.c |  138 +++++++++++++++++++++++++++++++++++++++++
 kernel/sched/sched.h           |    8 ++
 5 files changed, 190 insertions(+)

Index: linux-3.18.43-rt46/init/Kconfig
===================================================================
@ linux-3.18.43-rt46/init/Kconfig:420 @ config BSD_PROCESS_ACCT_V3
 	  for processing it. A preliminary version of these tools is available
 	  at <http://www.gnu.org/software/acct/>.
 
+config CPU_IDLERUNTIME
+	bool "Provide individual CPU usage measurement based on idle processing"
+	help
+	  If you say Y here, individual CPU usage data will be provided that are
+	  based on idle processing. The data are available for every CPU in
+	  /proc/cpuload/cpuN/data in the format "<idletime> <runtime>". The
+	  counters can be reset by writing to /proc/cpuload/resetall for all
+	  CPUs and to /proc/cpuload/cpuN/reset for a particular CPU,
+	  respectively. To calculate the usage since the most recent reset,
+	  the runtime must be devided by the sum of idletime plus runtime, e.g.
+	    cat /proc/cpuload/cpu0/data | awk '{print (100.0*$2) / ($1+$2)"%"}'
+
 config TASKSTATS
 	bool "Export task/process statistics through netlink"
 	depends on NET
Index: linux-3.18.43-rt46/kernel/sched/Makefile
===================================================================
--- linux-3.18.43-rt46.orig/kernel/sched/Makefile
+++ linux-3.18.43-rt46/kernel/sched/Makefile
@ linux-3.18.43-rt46/init/Kconfig:20 @ obj-y += wait.o wait-simple.o work-simpl
 obj-$(CONFIG_SMP) += cpupri.o cpudeadline.o
 obj-$(CONFIG_SCHED_AUTOGROUP) += auto_group.o
 obj-$(CONFIG_SCHEDSTATS) += stats.o
+obj-$(CONFIG_CPU_IDLERUNTIME) += cpu_idleruntime.o
 obj-$(CONFIG_SCHED_DEBUG) += debug.o
 obj-$(CONFIG_CGROUP_CPUACCT) += cpuacct.o
Index: linux-3.18.43-rt46/kernel/sched/core.c
===================================================================
--- linux-3.18.43-rt46.orig/kernel/sched/core.c
+++ linux-3.18.43-rt46/kernel/sched/core.c
@ linux-3.18.43-rt46/init/Kconfig:2362 @ prepare_task_switch(struct rq *rq, struc
 	fire_sched_out_preempt_notifiers(prev, next);
 	prepare_lock_switch(rq, next);
 	prepare_arch_switch(next);
+#ifdef CONFIG_CPU_IDLERUNTIME
+	if (is_idle_task(next)) {
+		int cpu = raw_smp_processor_id();
+
+		if (per_cpu(idlestop, cpu)) {
+			unsigned long flags;
+
+			raw_spin_lock_irqsave(&per_cpu(idleruntime_lock, cpu),
+			    flags);
+			per_cpu(idlestart, cpu) = cpu_clock(cpu);
+			per_cpu(runtime, cpu) +=
+			    per_cpu(idlestart, cpu) - per_cpu(idlestop, cpu);
+			raw_spin_unlock_irqrestore(&per_cpu(idleruntime_lock,
+			    cpu), flags);
+		}
+	} else if (is_idle_task(prev)) {
+		int cpu = raw_smp_processor_id();
+
+		if (per_cpu(idlestart, cpu)) {
+			unsigned long flags;
+
+			raw_spin_lock_irqsave(&per_cpu(idleruntime_lock, cpu),
+			    flags);
+			per_cpu(idlestop, cpu) = cpu_clock(cpu);
+			per_cpu(idletime, cpu) +=
+			    per_cpu(idlestop, cpu) - per_cpu(idlestart, cpu);
+			raw_spin_unlock_irqrestore(&per_cpu(idleruntime_lock,
+			    cpu), flags);
+		}
+	}
+#endif
 }
 
 /**
Index: linux-3.18.43-rt46/kernel/sched/cpu_idleruntime.c
===================================================================
--- /dev/null
+++ linux-3.18.43-rt46/kernel/sched/cpu_idleruntime.c
@ linux-3.18.43-rt46/init/Kconfig:4 @
+/*
+   cpu_idleruntime.c: provide CPU usage data based on idle processing
+
+   Copyright (C) 2012 Carsten Emde <C.Emde@osadl.org>
+
+   This program is free software; you can redistribute it and/or
+   modify it under the terms of the GNU General Public License
+   as published by the Free Software Foundation; either version 2
+   of the License, or (at your option) any later version.
+
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+
+   You should have received a copy of the GNU General Public License
+   along with this program; if not, write to the Free Software
+   Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA.
+*/
+
+#include <linux/seq_file.h>
+#include <linux/proc_fs.h>
+
+#include "sched.h"
+
+DEFINE_PER_CPU(unsigned long long, idlestart);
+DEFINE_PER_CPU(unsigned long long, idlestop);
+DEFINE_PER_CPU(unsigned long long, idletime);
+DEFINE_PER_CPU(unsigned long long, runtime);
+DEFINE_PER_CPU(raw_spinlock_t, idleruntime_lock);
+
+static int idleruntime_show(struct seq_file *m, void *v)
+{
+	unsigned long cpu = (unsigned long) m->private;
+	unsigned long long now;
+	unsigned long flags;
+
+	raw_spin_lock_irqsave(&per_cpu(idleruntime_lock, cpu), flags);
+
+	/* Update runtime counter */
+	now = cpu_clock(cpu);
+	per_cpu(runtime, cpu) += now - per_cpu(idlestop, cpu);
+	per_cpu(idlestop, cpu) = now;
+
+	seq_printf(m, "%llu %llu\n", per_cpu(idletime, cpu),
+	    per_cpu(runtime, cpu));
+
+	raw_spin_unlock_irqrestore(&per_cpu(idleruntime_lock, cpu), flags);
+
+	return 0;
+}
+
+static inline void idleruntime_reset1(int cpu)
+{
+	unsigned long flags;
+
+	raw_spin_lock_irqsave(&per_cpu(idleruntime_lock, cpu), flags);
+	per_cpu(idletime, cpu) = per_cpu(runtime, cpu) = 0;
+	per_cpu(idlestop, cpu) = cpu_clock(cpu);
+	raw_spin_unlock_irqrestore(&per_cpu(idleruntime_lock, cpu), flags);
+}
+
+static ssize_t idleruntime_reset(struct file *file, const char __user *buffer,
+				 size_t len, loff_t *offset)
+{
+	unsigned long cpu = (unsigned long) PDE_DATA(file_inode(file));
+
+	idleruntime_reset1(cpu);
+	return len;
+}
+
+static ssize_t idleruntime_resetall(struct file *file,
+				    const char __user *buffer,
+				    size_t len, loff_t *offset)
+{
+	unsigned long cpu;
+
+	for_each_online_cpu(cpu)
+		idleruntime_reset1(cpu);
+	return len;
+}
+
+static const struct file_operations idleruntime_resetall_fops = {
+	.write = idleruntime_resetall,
+	.release = single_release,
+};
+
+static int idleruntime_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, idleruntime_show, PDE_DATA(inode));
+}
+
+static const struct file_operations idleruntime_fops = {
+	.open = idleruntime_open,
+	.read = seq_read,
+	.llseek = seq_lseek,
+	.write = idleruntime_reset,
+	.release = single_release,
+};
+
+static int __init proc_idleruntime_init(void)
+{
+	unsigned long cpu;
+	struct proc_dir_entry *root_idleruntime_dir;
+
+	root_idleruntime_dir = proc_mkdir("idleruntime", NULL);
+	if (!root_idleruntime_dir)
+		return 0;
+
+	if (!proc_create("resetall", S_IWUGO, root_idleruntime_dir,
+	    &idleruntime_resetall_fops))
+		return 0;
+
+	for_each_possible_cpu(cpu) {
+		char name[32];
+		struct proc_dir_entry *idleruntime_cpudir;
+
+		raw_spin_lock_init(&per_cpu(idleruntime_lock, cpu));
+
+		snprintf(name, sizeof(name), "cpu%lu", cpu);
+		idleruntime_cpudir = proc_mkdir(name, root_idleruntime_dir);
+		if (!idleruntime_cpudir)
+			return 0;
+
+		if (!proc_create_data("data", S_IRUGO, idleruntime_cpudir,
+		    &idleruntime_fops, (void *) cpu))
+			return 0;
+
+		if (!proc_create_data("reset", S_IWUGO, idleruntime_cpudir,
+		    &idleruntime_fops, (void *) cpu)) {
+			remove_proc_entry("data", idleruntime_cpudir);
+			return 0;
+		}
+	}
+	return 0;
+}
+
+module_init(proc_idleruntime_init);
Index: linux-3.18.43-rt46/kernel/sched/sched.h
===================================================================
--- linux-3.18.43-rt46.orig/kernel/sched/sched.h
+++ linux-3.18.43-rt46/kernel/sched/sched.h
@ linux-3.18.43-rt46/init/Kconfig:681 @ static inline u64 rq_clock_task(struct r
 	return rq->clock_task;
 }
 
+#ifdef CONFIG_CPU_IDLERUNTIME
+extern DEFINE_PER_CPU(unsigned long long, idlestart);
+extern DEFINE_PER_CPU(unsigned long long, idlestop);
+extern DEFINE_PER_CPU(unsigned long long, idletime);
+extern DEFINE_PER_CPU(unsigned long long, runtime);
+extern DEFINE_PER_CPU(raw_spinlock_t, idleruntime_lock);
+#endif
+
 #ifdef CONFIG_NUMA_BALANCING
 extern void sched_setnuma(struct task_struct *p, int node);
 extern int migrate_task_to(struct task_struct *p, int cpu);