Re: Time to fork the netboot code?

[Mario Klebsch]

-----BEGIN PGP SIGNED MESSAGE-----

On Wednesday,  9. May 2001 21:19, Chip Salzenberg wrote:
> According to Mario Klebsch:
> > The goal would be to convert the situation from
> >     the grub people using the drivers of the etherboot people
> > to
> >     the grub and etherboot people are using common drivers.
>
> If the Etherboot people would be happy with it, that would be an
> improvement.  But I think it's inevitable that we'd be behind the
> state of the art until we started using the driver set from an OS
> that's actually maintained, like Linux or one of the BSDs.

The idea to have common drivers with a main stream OS looks good, at
first glance. But I am not convinced, that it really is.

The reqirements for drivers of real estate OS are much different that
our requirements.

Linux drivers have to be able to hande multicasts and lots of other
stuff (ioctls, ...). Theses drivers have to use interupts and have to
work on MP systems, they have bottom halves, thes uses the systems
memory management and often they have to be thread safe.

OTOH, tghe drivers for booting have to be simple. None of the etherboot
drivers uses interupts for the NIC, and the use of DMA leads to OS
crashes, the the DMA channels are not deactivated prior to booting the
OS.

Allthough the requirements differ, I am sure, it is possible to have
both drivers in one source file, when using advanced preprocessor
tricks. But this would at least require, that both driver flavours are
maintained by the same set of people. If not, we will face the same
problem, that we have today, with the etherboot drivers.

When I see, that even the Linux and the *BSD drivers do not use a
common source file, I really cannot imagine GRUB and Linux (or *BSD)
cooperating this was in THIS world. :-(

So my suggestion would be to define a 'non-OS' network driver API,
which is usable for all bootstrap projects, agree with these projects
upon this API and then convert the existing drivers to the new API.

Of course, to follow the other aproach (a common OS/bootstrap driver),
we could get in touch with the linux/*BSD network driver people and ask
then, what they do think about this idea? They know their code much
better than I do, and perhaps they like this idea. Despite all my
doubts, why should I not be an optimist? :-)

73, Mario
- --
Mario Klebsch                                           address@hidden
PGP-Key available at http://www.klebsch.de/public.key
Fingerprint DSS: EE7C DBCC D9C8 5DC1 D4DB  1483 30CE 9FB2 A047 9CE0
 Diffie-Hellman: D447 4ED6 8A10 2C65 C5E5  8B98 9464 53FF 9382 F518


-----BEGIN PGP SIGNATURE-----
Version: PGPfreeware 5.0i for non-commercial use
MessageID: QIb5LND0zIFtl8cxwNZL1rc4z/jJsFHw

iQA/AwUBOvo/+TDOn7KgR5zgEQINzACeJQ4nI/EqO0tGmlw/yRU07bNIFj8AoOXI
8Vz+AJUSl767voawmhYRBpt8
=isP7
-----END PGP SIGNATURE-----