Jtdylktuy

Linux's /proc/PID/stat lists several metrics that are measured in jiffies. Is there a way to get the current # of jiffies since reboot? Jiffies I also assume to be seconds times the USER_HZ value.

Can I get this by summing the first 4 arguments on the cpu line in /proc/stat?

UPDATE:

$ date +"%s.%N" && grep '^jiffies' timer_list

1262103103.162169230

jiffies: 1007865965

jiffies: 1007865965

jiffies: 1007865965

jiffies: 1007865965

$ date +"%s.%N" && grep '^jiffies' timer_list

1262103108.706475051

jiffies: 1007867351

jiffies: 1007867351

jiffies: 1007867351

jiffies: 1007867351

The delta between these is almost exactly 250 jiffies / second. Appears as though all the jiffies lines per CPU are equivalent (though I guess they might not be if a hot | ondemand CPU was provisioned/replaced). I think this gives me the best place to pull a consistent jiffies counter.

Technically jiffy in computer parlance is the duration of 1 tick of the system timer interrupt. It's not absolute though. For Linux 2.6.13+ on Intel x86 jiffy is 4ms, but can range from 1ms to 10ms depending upon architecture and kernel version.

From the Kernel Timer Systems page:

Historically, the kernel used 100 as the value for HZ, yielding a jiffy interval of 10 ms. With 2.4, the HZ value for i386 was changed to 1000, yeilding a jiffy interval of 1 ms. Recently (2.6.13) the kernel changed HZ for i386 to 250. (1000 was deemed too high).

It lists /proc/timer_list and /proc/timer_stats.

You can activate the timer_stats at boot time, then cat this file to print stats.

No, you only need the first line. The first line aggregates everything else in the other cpu lines.

Example output:

[john@awesome]$cat /proc/stat

cpu  35024984 1771325 94153391 1810948613 2648063 352387 557232

cpu0 13955475 927654 59431476 895791946 1910028 318618 438048

cpu1 21069509 843671 34721915 915156667 738035 33769 119184

intr 1403502159 1138402452 597 0 3 3 0 5 0 1 0 0 0 12315 0 92119425 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 57676632 0 0 0 0 0 0 0 115290726 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

ctxt 21043582666

btime 1252332786

processes 25663823

procs_running 1

procs_blocked 0

What each column means (left to right):

• user: normal processes executing in user mode


• nice: niced processes executing in user mode


• system: processes executing in kernel mode


• idle: idle time


• iowait: waiting for I/O to complete


• irq: servicing interrupts


• softirq: servicing softirqs

as you can see, the first column after cpu (user mode processes) is equal to the 2 numbers beneath it added together.

jiffies per second:

awk 'BEGIN {"cat /proc/timer_list | grep '''^jiffies''' | awk '''{print $2}'''" | getline a; "cat /proc/uptime | awk '''{print $1}'''" | getline b ;printf "%.4fn", a/b}'

Explanation: It divides jiffies since boot found in /proc/timer_list by the seconds since boot found in /proc/uptime

In later Linux Kernels you can query /proc/timer_list and find out the number of jiffies since reboot from each CPU that's present within the system. They should always match.

$ grep -E "^cpu|^jiff" /proc/timer_list

cpu: 0

jiffies: 4299690231

cpu: 1

jiffies: 4299690231

If you take a look at the code behind timer_list.c the bit that prints the above:

...

...

        P_ns(iowait_sleeptime);

        P(last_jiffies);

        P(next_timer);

        P_ns(idle_expires);

        SEQ_printf(m, "jiffies: %Lun",

               (unsigned long long)jiffies);

Notice that it's unsigned long long.