What is user mode jiffies and kernel mode jiffies of a process

iamjayanth · 09-14-2009, 04:49 AM

Hi all,

I am trying to learn /proc/<pid>/stat file contents. And when I come across two labels

user mode jiffies
kernel mode jiffies

I try to google for jiffies, and I got this data.

Quote:

An incrementing counter representing system "uptime" in ticks - or the number of timer interrupts since boot. Ultimately the entire original concept of a jiffy will likely vanish as systems use timer events only when necessary and become "jiffyless".

What this tells me is that jiffies is a global variable that starts from zero and increments in each clock intervals(cycle).I didnt understand the second sentense in that quote.

Also why process have user mode jiffies and kernel mode jiffies. if these times tell how many times this process executed in user space and in kernel space(system calls) respectively ? Please help

lutusp · 09-14-2009, 05:26 AM

Quote:

Originally Posted by iamjayanth

Hi all,

I am trying to learn /proc/<pid>/stat file contents. And when I come across two labels

user mode jiffies
kernel mode jiffies

I try to google for jiffies, and I got this data.

What this tells me is that jiffies is a global variable that starts from zero and increments in each clock intervals(cycle).I didnt understand the second sentense in that quote.

Also why process have user mode jiffies and kernel mode jiffies. if these times tell how many times this process executed in user space and in kernel space(system calls) respectively ? Please help

The reason jiffies will go away is because the original way the kernel used to slice up time for processes is changing. Processes will no longer get a time slice unless they actually need it, which renders the jiffie scheme meaningless.

"Multitasking" once meant an arbitrary division of processor time between all running processes. Now the kernel scheduler can avoid assigning time to a process that doesn't actually need the time.

Jiffies are separated into kernel and user space because system administrators want to know how much time is spent in kernel routines versus user-space code.

iamjayanth · 09-14-2009, 11:35 PM

Thanks lutusp for the reply...

So you are saying that at present scheduler allocates a time slice for each and every process beforehand, without checking whether they need that or not...And scheduler allocates this time in units of jiffies...
And in future(?) each and every process need to request for there exicution time, which a scheduler can provide.

Then pre calculation of time for processes in terms of jiffies will go making jiffies meaningleass....

Am I understood correctly...

Sorry if my english is wrong...