Bug 66249

Summary:	Constant 1.00 loadavg without apparent cause
Product:	Base System	Reporter:	Krzysztof Kowalik <kkowalik+pr>
Component:	i386	Assignee:	freebsd-i386 (Nobody) <i386>
Status:	Closed FIXED
Severity:	Affects Only Me
Priority:	Normal
Version:	Unspecified
Hardware:	Any
OS:	Any

Description Krzysztof Kowalik 2004-05-04 12:50:28 UTC

    constant 1.00 loadavg, with top saying: 

CPU states:  0.0% user,  0.0% nice,  0.0% system,  1.6% interrupt, 98.4% 
idle 

(With both SMP and non-SMP kernels, with the default scheduler)

After recompiling the kernel with SHED_ULE, loadavg dropped to the constant 0.99, so it still does not exactly look right.

  PID USERNAME PRI NICE   SIZE    RES STATE    TIME   WCPU    CPU COMMAND
   11 root     -16    0     0K    12K RUN      8:48 97.17% 97.17% idle
   27 root     -44 -163     0K    12K WAIT     0:07  0.00%  0.00% swi1: net
   28 root     -48 -167     0K    12K WAIT     0:01  0.00%  0.00% swi8: tty:sio clock
[...]

And kern/21155 pr does not exactly include the possible fix (nor hints).

How-To-Repeat: to reboot and wait, nothing more.

Comment 1 Krzysztof Kowalik 2004-05-14 21:36:04 UTC

After some blind-hits:

# sysctl kern.polling.enable=0
kern.polling.enable: 1 -> 0

And load average immediately went down.

# uptime
22:34  up 10 days, 9 hrs, 1 user, load averages: 0,05 0,28 0,63

-- 
Krzysztof Kowalik, Computer Center, AGH UST.

Comment 2 Maxim Konovalov freebsd_committer

2004-07-17 23:49:20 UTC

State Changed
From-To: open->closed

Not a bug, polling(4) is a culprit.