Subject: Re: nfsd and process load sharing
To: None <>
From: Guenther Grau <>
List: tech-kern
Date: 06/04/1999 14:50:40
Frank van der Linden wrote:
> On Fri, Jun 04, 1999 at 09:28:04PM +1000, matthew green wrote:
> >
> > actually, this would mean that on a lightly loaded system, every
> > nfsd would be kept in memory as they would round robin, rather
> > than just having the few necessary in core, and the others possible
> > paged out...
> >
> > i think i prefer the current behaviour..
> Well, I wasn't serious about it to begin with. Though it would not

Then your mail was missing the appropriate ;-) to indicate
that this was meant ironically/not seriously :-)

> have mattered much, nfsds spend their lives almost entirely in the
> kernel, in memory that will always be paged in.

I didn't know this. Thanx for the enlightment. I'll have
to go digging through the code to see how a process can
enter the kernel and spend most of it's time there :-)

Besides that, it might be possible that we make parts
of the kernel memory pageable. I thought pool'ifying 
the memory allocations was the first step towards this,
not (besides making memory allocations more dynamically