Subject: Re: LKMs
To: =?X-UNKNOWN?Q?Jarom=EDr_Dolecek?= <dolecek@ics.muni.cz>
From: Al Snell <alaric@alaric-snell.com>
List: tech-kern
Date: 11/25/2000 19:42:24
On Sat, 25 Nov 2000, Jarom=EDr Dolecek wrote:

> > That sounds good.. is there going to be a securelevel where even
> > 'registered' LKM's can't be loaded?
>=20
> Depends on how usable this would be. It's too early to tell now,
> anyway.

Reminds me - my initial thinking was that it'd be easier to NOT provide
standard monolothic kernel support, BUT to make use of the fact that some
means of providing a set of modules to the initial kernel image needs to
exist in the boot loader to set up a kernel that never has the ability to
load new modules... the effect is the same, but it saves from needing two
seperate ways of configuring the kernel, since most of the hardware setup
stuff in a kernel config file will need to be doable at run time anyway,
might as well ditch the old thing.

I was thinking that the initial kernel image could use a variant of that
thing the install disks use where a disk image is sorta compiled in. It'd
get mounted under /boot.lkm instead of being the root device, and would be
read only. The highest-securelevel kernel would refuse to accept module
registrations, but all in /boot.lkm are registered during the boot
process. security level 0 would allow the registration of new modules. The
process of "loading" a module would exist in all securelevels.

The remaining sliver of a monolothic kernel would need the
disk-image-in-kernel-file thing in it as the only compiled-in filesystem.

I've rambled lots so I'll shut up now.

ABS


--=20
                               Alaric B. Snell
 http://www.alaric-snell.com/  http://RFC.net/  http://www.warhead.org.uk/
   Any sufficiently advanced technology can be emulated in software =20