Native preemption is the culprit [was Re: today's CURRENTlockups]

[Daniel Eriksson]

Robert Watson wrote:

> Per my earlier post, I've experienced similar unreliability.  The work
> around I'm using is to build with "#define PREEMPTION" from param.h
> disabled.  This results in a quite usable kernel, although 
> quite probably
> more poor interrupt handling latency, etc.

The server I'm running with preemption turned off becomes irresponsive and
the load average skyrockets (10+) whenever I do something as simple as
bzip2-compressing a file. Setting a nice value of 20 does not help at all.
Are you experiencing similar results on your machine?

/Daniel Eriksson