Robert R. Russell
zoohoo.linux at gmail.com
Wed Jan 7 23:53:37 PST 2009
On Wednesday 07 January 2009 05:54:01 pm Ken Moffat wrote:
> Just a reminder that making the right decisions when new choices
> appear in .config is not always obvious and needs to be tested.
> With 2.6.27 I was happy that everything worked adequately. Tried
> several 2.6.28-rc and made what seemed to be sensible choices. At
> that time I wasn't building a new system. When 2.6.28 came out I
> rolled it out across my boxes and systems (several on each box).
> And then I built a new pure64 system. In the initial stages I
> noted that firefox was lagging horribly during compilations (e.g.
> start it during a compile, or open a new tab, and nothing seems to
> happen for far longer than 'normal', and the system was as
> unresponsive as I've ever seen.
> Compared the kernel options, the only obvious change was that for
> some reason I'd selected CONFIG_GROUP_SCHED (in the past, I had
> different problems with that on x86_32). Turned that off for the
> new system, and (now I've built nearly everything) responsiveness is
> back to how I expect.
> Another one to add to the bulky file labelled "experience", but I
> do wonder what sort of hardware the people who use group scheduling
> are running on.
> das eine Mal als Tragödie, das andere Mal als Farce
If what I have been told by some others is correct, group scheduling only
works if certain userspace utilities are installed a configured. I don't know
what utilities are required though.
More information about the lfs-chat