home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Turnbull China Bikeride
/
Turnbull China Bikeride - Disc 2.iso
/
BARNET
/
ARMLINUX
/
MAIL
/
9804
/
000119_9606585c@udcf.gla.ac.uk _Tue Apr 21 10:04:25 1998.msg
< prev
next >
Wrap
Internet Message Format
|
1998-05-13
|
5KB
Return-Path: <9606585c@udcf.gla.ac.uk>
Received: from lenzie.cent.gla.ac.uk (lenzie.cent.gla.ac.uk [130.209.30.11])
by odie.barnet.ac.uk (8.8.6/8.8.6) with ESMTP id KAA22884
for <willy@odie.barnet.ac.uk>; Tue, 21 Apr 1998 10:04:23 +0100
Received: from localhost (9606585c@localhost)
by lenzie.cent.gla.ac.uk (8.8.4/8.8.8) with SMTP id KAA16097;
Tue, 21 Apr 1998 10:04:37 +0100 (BST)
Date: Tue, 21 Apr 1998 10:04:34 +0100 (BST)
From: James Craig <9606585c@udcf.gla.ac.uk>
X-Sender: 9606585c@lenzie.cent.gla.ac.uk
To: Matthew Wilcox <willy@odie.barnet.ac.uk>
cc: linux-arm@vger.rutgers.edu
Subject: Re: Installing on an 4 MB A5000
In-Reply-To: <199804210330.EAA21711@odie.barnet.ac.uk>
Message-ID: <Pine.GSO.3.95.980421095758.7028E-100000@lenzie.cent.gla.ac.uk>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Status: RO
On Tue, 21 Apr 1998, Matthew Wilcox wrote:
> Russell King - ARM Linux Admin
> >
> > Hi.
> >
> > Maybe I ought to make my position quite clear on this subject:
> >
> > 1) I believe that RedHat offers the best compromise for ARM Linux.
>
> That's your belief and you're entitled to hold it. Since you created the
> distribution, others should accept that. I still like the idea of creating
> a Debian distribution, but I think that time spent working on that would
> be wasted before we have libc 6.
Possibly, although I'm not sure. I can't think of any reason why slackware
or Red Hat should be any different in that respect. :)
> > 2) The RedHat installer cannot be used on floppy to install on machines
> > with 4MB of ram. With a future PartMan, it will be possible to copy
> > a small partition image into the drive, thus allowing non-floppy
> > based installation.
>
> That will be really sexy.
Yep, very nice idea. That would solve a *lot* of problems with the setup.
> > 4) The 2.1.xx kernels currently are going to require a lot of changes
> > to get them to run properly on the older architectures:
> > a) The MEMC is quite unlike any other memory manager in the world,
> > and doesn't fit any sensible memory management model that Linux
> > could provide. Therefore, the MM is less than optimal.
>
> You can say that again!
Yep. :)
> > b) With the advent of ELF, and it's requirement to map the same
> > data in two logical locations, this will require more
> > modifications to the kernel.
>
> I *thought* it had been agreed that that wasn't actually necessary, it was
> merely the way the x86 decided to do it?
Likewise, but hey, I don't understand ELF all that well anyway. :)
> > c) The large page size requires numerous modifications to the
> > way memory is allocated in several parts of Linux.
>
> Yes, I noticed that in the patches. Are there not also issues on Alpha's
> with their 8k page sizes? I guess they tend to have so much more memory
> that this is a moot point.
Truly.
> > d) The kernel's malloc() routines need a hack to reduce the
> > inherent overheads that it produces on large page machines.
> > I'm not saying that it'll be impossible to do, but more that
> > I'm beginning to wonder if the old machines are really worth the
> > effort, and whether we're going to be able to continue integrating
> > the architecture-specifics for these machines with obsolete
> > processors into Linus' kernel source tree.
Leaving support for large page sizes in the kernel would be a good move,
since it increases the range of different CPUs that linux *could* be
ported to with relatively little work.
> > 5) It is my intention to continue integrating the sources as is or as
> > patches allow for these old machines. I shall be working on this
> > area, but since my resources are required for other purposes, this
> > will not happen with any great speed. If someone wishes to take
> > over the kernel admin/hacking for the A5000, they are welcome. If
> > you want to know more details about this, then please mail me.
I'm in no position to help at present, sadly. :(
> IMO, it would make more sense to separate out the port into an arm26 and
> an arm32. Yeah, I know this is a really bad time to mention it :-) Of
> course, there's nothing to stop arm32 machines executing arm26 binaries.
> This might encourage someone to take responsibility for the arm26 tree
> off your hands.
Interesting idea, but I suspect that it would lead to a complete halt on
progress on the arm26 architecture. At least with the two together, we
still get progress - if arm26 and arm32 were seperate, we'd have to update
the arm26 code seperately from the arm32 stuff, even if the changes were
applicable to both architectures.
--
James Craig <jcraig@mad.scientist.com>
<9606585c@udcf.gla.ac.uk>