[ On Sun, February 15, 1998 at 13:08:57 (-0800), Jason Thorpe wrote: ]
> Subject: Re: UVM for sun3? 
>
> UVM hasn't been ported to the sun3 yet.  I've done the m68k generic bits
> which I will commit shortly (since I almost have it working on the hp300),
> but some changes will have to be made to the sun3 port to make it work.

That shouldn't be a problem, esp. now that I know you've got the CPU
dependent bits fixed up....  I think I can find and fix other
dependencies by inspection of one of the ports that does work already.
(esp. if I wait for the hp300 commits!  ;-)

>  > Speaking of user-land, how close are we to having 'ps', 'w', etc.
>  > working with UVM?
> 
> Works... put "UVM=yes" in /etc/mk.conf, and rebuild the world.

I would if I could, but until I get just the minimum necessary bits of
source I'm kinda stuck.

BTW, this is one of those things that makes me long for the elegant way
SunOS-4 keeps kernel-arch and user-arch stuff in separate parts of the
filesystem.  Then I could have one kernel-arch filesystem where the
necessary dependent bits were separated out, one for each kernel
specific version.  I.e. a specific /usr/kvm with 'ps', 'w', libkvm*, et
al for each major variant of a given kernel such as 'VM' vs. 'UVM',
"-current" vs. "release", etc.  A minor tweak of a symlink or mount
would then suffice to adjust user-land for the current kernel, and
running diskless systems with different kernel-arch variants would be
far more efficient.

-- 
							Greg A. Woods

+1 416 443-1734			VE3TCP			robohack!woods
Planix, Inc. <woods@planix.com>; Secrets Of The Weird <woods@weird.com>