Question about scheduling in 2.4.20

Question about scheduling in 2.4.20

[Ravinandan Arakali (rarakali)]

Hi,
I have a question about scheduling latencies in 2.4.20. We are seeing
that our
process (started with default priority, normal fork, exec) is not
scheduled
on the CPU for 4 seconds (from our kernel traces). Is that possible
under
heavy load ? We believe the process is still alive but I wanted to know
if
anybody has an idea from experimental results, about the worst case
scenario.
>From syscall trace, we also know that the process is not inside any
blocking
system call.

Thanks,
Ravi 

RE: Question about scheduling in 2.4.20

[Ravinandan Arakali (rarakali)]

Hi,
I notice with our version, it's running 2.4.20 but has the O(1)
scheduling patch.

Specifically, we see that the process sleeps inside an ioctl for more
than 3 secs, 
wakes up, comes out of the blocking call and is immediately scheduled
out. After 
this, it's not scheduled for next 4 secs. 
Does this look like a fairness/starvation issue ?

Thanks,
Ravi 

-----Original Message-----
From: Ravinandan Arakali (rarakali) 
Sent: Thursday, June 14, 2007 10:02 AM
To: Linux Kernel
Subject: Question about scheduling in 2.4.20

Hi,
I have a question about scheduling latencies in 2.4.20. We are seeing
that our process (started with default priority, normal fork, exec) is
not scheduled on the CPU for 4 seconds (from our kernel traces). Is that
possible under heavy load ? We believe the process is still alive but I
wanted to know if anybody has an idea from experimental results, about
the worst case scenario.
>From syscall trace, we also know that the process is not inside any
blocking system call.

Thanks,
Ravi