Re: booting, bochs l4ka

[Matthew Sackman]

On Fri, May 17, 2002 at 02:41:02PM +0200, Farid Hajji wrote:
> Right. L4ka/Hazelnut is the current microkernel, but we plan to
> write l4hurd on top of L4ka/Pistachio (X.2, later Version 4).
> But Pistachio is not yet finished/released. In the mean time,
> we use Hazelnut though.

Yep, I'd got all of that. Thanks for clarifying. From having been on the
mailing list for a month or two, people are just using this time to get
some experience with l4-ka and start planning how to get hurd on to it?

> I tried L4Ka/Hazelnut (cvs version 2 days old) with
>   gcc version 2.95.3 20010315 (release) [FreeBSD]
> and it works just fine. gcc 2.95.4 up to gcc 3.0.3 are known to
> be broken and they generate corrupt binaries.

Currently compiling gcc 3.1.

> The 4MB superpages patch is still present in the CVS version of Bochs,
> but no-one volunteered to merge it in the main sources right now. Since
> I'm currently using a physical pentium for tests, I simply didn't have
> the incentive to update the patch :) Please feel free to contribute here.

Ok, I'll look at that asap. I do have A-levels over the next month so
I'm only going to hack this when I really do have free time (normally I
just alter my life to the job at hand, but alas I really can't do that
now!).

A quick look shows that it doesn't apply at all. Agh. Looks like a long
slow manual patch and rewrites will be needed... will know more when
I've had a detailed look. Any pointers you can give would be great.

> The patch to l4-ka is certainly out of date. That was just a quick
> hack to get going.

What does that patch actually do and why is it needed? Is it simply to
convince l4-ka that "it's OK" to boot under bochs? Thus should it
actually be integrated into l4-ka, or is it to address a problem with
bochs (in which case should it be integrated into the bochs patch)?
Quick look seems to show that it's updating the asm code (starting up
the kernel?) and altering the interrupt code (make l4-ka happy with the
'hardware' that bochs provides?). So therefore should it be a compile
option?

> I'm very interested in an updated Bochs with 4MB superpages that
> could boot current Hazelnut. Bochs is incredibly useful to trace
> single instructions and memmory contents and is IMO easier to use
> than gdb-over-serial-line :). It's not only interesting to replace
> missing hardware, but _is_ useful as a fine debugging too as well.

I don't doubt it, I really detest boot up times and multiple reboots, so
anything that can help there I'm happy to help with. I've read through
the patch for the 4MB superpages and obviously l4-ka needs this to
function. Can one ask why? It just seems a little odd that this is such
a rare requirement that the developers of Bochs don't consider it
necessary to integrate into bochs. So why does l4-ka need this, but
everything else that bochs can deal with doesn't?

> Anyone volunteering to take over the 4MB superpages patch and help
> integrate it seamlessly in the current Bochs sources? I'm very short
> on time right now, but I'm willing to assist any brave soul that
> wants to step forward here.

Brave soul? Oh dear, may be out of my reach then. I'll give it a shot,
but don't expect anything for at least a month or two!

(and expect a few questions along the way: my assembler coding never
went much further than the BBC micro (6502: 2 registers + accumulator!))

Many thanks for your helpful reply.

Matthew

-- 

Matthew Sackman
Nottingham
England

BOFH Excuse Board:
Root name servers corrupted.