OSEC

Neohapsis is currently accepting applications for employment. For more information, please visit our website www.neohapsis.com or email hr@neohapsis.com
 
Re: modules on the CD - yet another question

From: Rafal Wojtczuk (rafal.wojtczuk7bulls.com)
Date: Thu Mar 11 2004 - 16:45:44 CST

On Fri, Mar 12, 2004 at 01:13:53AM +0300, Solar Designer wrote:
> > Marcin Siennicki wrote to owl-userslists.openwall.com:
> > > this time I'm asking why there are no device drivers
> > > (for SCSI controllers, NICs,...) on the CD?
[cut]
> We already had to drop some potentially useful stuff from default
> kernels on the CDs (most notably, PCMCIA and USB support) when moving
> from 2.2.x to the bigger 2.4.x kernels, also adding support for ext3.
[cut]
> A better approach might be to move all the disk controller drivers
> into the 1.44 MB floppy image (either compile them into the kernel
> like we do now or start using initrd), but to make that possible at
> the same time remove all non-essential non-disk drivers and even
> entire kernel subsystems into modules available only on the CD (not in
Well, perhaps the long-time goal would be to support initrd, and during the
installation to generate a proper initrd with required drivers (like
"mkinitrd"(8) does on RH). This require significant developer time, though.

> There's really no room to add many more drivers with the current
> approach. Well, maybe if we switch to using 2.88 MB "floppies", but
> that would kill support for systems which can't boot off a CD and
> actually require one to produce a boot floppy.
I'd like to keep the ability to boot from floppy, it is useful in many
situations.

> > some (conflicting like 3COM 3c5x9 and 3c90x)
>
> I'm not sure you're correct about this pair of drivers. CONFIG_EL3
> (for 3c509) and CONFIG_VORTEX (for 3c59x and 3c90x) may both be
> enabled at the same time, -- and are enabled on the Owl 1.1 CDs.
They can be both compiled in the kernel - which is in fact a bug in kernel
config. I once experienced a situation when 3c90x driver claimed the
resources of a 3c59x card (because PCI IDs are similar enough), which resulted
in non-functional eth0.

> There're many good things to do...
So much to do, so little time - sounds like a quotation, but I don't
remember the source :)

RW