Question

There are a number of posts and references on how to get CPU Utilization using statistics in /proc/stat. However, most of them use only four of the 7+ CPU stats (user, nice, system, and idle), ignoring the remaining jiffie CPU counts present in Linux 2.6 (iowait, irq, softirq).

As an example, see Determining CPU utilization.

My question is this: Are the iowait/irq/softirq numbers also counted in one of the first four numbers (user/nice/system/idle)? In other words, does the total jiffie count equal the sum of the first four stats? Or, is the total jiffie count equal to the sum of all 7 stats? If the latter is true, then a CPU utilization formula should take all of the numbers into account, like this:

#include <stdio.h>
#include <stdlib.h>

int main(void)
{
  long double a[7],b[7],loadavg;
  FILE *fp;

  for(;;)
  {
    fp = fopen("/proc/stat","r");
    fscanf(fp,"%*s %Lf %Lf %Lf %Lf",&a[0],&a[1],&a[2],&a[3],&a[4],&a[5],&a[6]);
    fclose(fp);
    sleep(1);
    fp = fopen("/proc/stat","r");
    fscanf(fp,"%*s %Lf %Lf %Lf %Lf",&b[0],&b[1],&b[2],&b[3],&b[4],&b[5],&b[6]);
    fclose(fp);

    loadavg = ((b[0]+b[1]+b[2]+b[4]+b[5]+b[6]) - (a[0]+a[1]+a[2]+a[4]+a[5]+a[6]))
         / ((b[0]+b[1]+b[2]+b[3]+b[4]+b[5]+b[6]) - (a[0]+a[1]+a[2]+a[3]+a[4]+a[5]+a[6]));
    printf("The current CPU utilization is : %Lf\n",loadavg);

  }

  return(0);
}

Answer 1

I think iowait/irq/softirq are not counted in one of the first 4 numbers. You can see the comment of irqtime_account_process_tick in kernel code for more detail:

(for Linux kernel 4.1.1)

2815  * Tick demultiplexing follows the order
2816  * - pending hardirq update    <-- this is irq
2817  * - pending softirq update    <-- this is softirq
2818  * - user_time
2819  * - idle_time         <-- iowait is included in here, discuss below
2820  * - system time
2821  *   - check for guest_time
2822  *   - else account as system_time

For the idle time handling, see account_idle_time function:

2772 /*
2773  * Account for idle time.
2774  * @cputime: the cpu time spent in idle wait
2775  */
2776 void account_idle_time(cputime_t cputime)
2777 {
2778         u64 *cpustat = kcpustat_this_cpu->cpustat;
2779         struct rq *rq = this_rq();
2780
2781         if (atomic_read(&rq->nr_iowait) > 0)
2782                 cpustat[CPUTIME_IOWAIT] += (__force u64) cputime;
2783         else
2784                 cpustat[CPUTIME_IDLE] += (__force u64) cputime;
2785 }

If the cpu is idle AND there is some IO pending, it will count the time in CPUTIME_IOWAIT. Otherwise, it is count in CPUTIME_IDLE.

To conclude, I think the jiffies in irq/softirq should be counted as “busy” for cpu because it was actually handling some IRQ or soft IRQ. On the other hand, the jiffies in “iowait” should be counted as “idle” for cpu because it was not doing something but waiting for a pending IO to happen.