Subject: Re: kernel_pt_table should be static?
To: John Fremlin <vii@users.sourceforge.net>
From: Chris Gilbert <chris@paradox.demon.co.uk>
List: port-arm
Date: 05/27/2001 01:18:10
On Saturday 26 May 2001 11:40 pm, John Fremlin wrote:
> Chris Gilbert <chris@paradox.demon.co.uk> writes:
> > On Saturday 26 May 2001 9:44 pm, John Fremlin wrote:
> > > This variable isn't used outside of the individual *_machdep.c where
> > > it is defined?
> >
> > Could be, at one time it might have been used outside of the machdep
> > files, only way to find out is to make it static and see if your
> > kernel still links. It it works please send us the diff and someone
> > will commit it. :)
>
> That test passed before I sent the mail - the trouble is that my
> kernel is the first try at a new port.
Excellent stuff :)
Currently I'm working on the pmap to figure out where I can improve things.
Looking at the i386 pmap it's got a lot of useful things/ideas that may well
appear in the arm32 pmap. It seems that the rewrite of it for UVM may have
done wonders for it, I know that the arm32 pmap was based on the i386 pmap
from years back, and mark said it was the wrong choice, now it appears that
it might be the right choice. However first I'm going to track down why we
take so long in a fork exit process
> > It's probably not the only bit of cruft you'll find lying around the
> > place that should be garbage collected.
>
> Indeed. Indeed. I'm rewriting initarm now (a spot of much needed work
> IMHO). Would other ARMs want it? ATM it is just for psionw.
The thing is that every arm arch has it's own version of initarm, certainly
some features should be portable to others though, so certainly mail around
about any updates/things you find, see if we can use them elsewhere :)
Cheers,
Chris