Re: Subsystem booting

[Marcus Brinkmann]

On Sun, May 25, 2003 at 05:22:15PM +0200, Niels Möller wrote:
> Sure, but for each subsystem there will only be one task (or rather, a
> small number of threads) which the rootserver consideres privileged.

That's the current plan but it is not a solid rule.

> It seems simpler if that corresponds to one grub module, but you're
> right that's not absolutely necessary.

Again, these are issues that we can and do not need to fix at the current time.
 
> BTW, I spent the afternoon rereading the L4 spec. Section 7.7 says
> that the kernel can start three special tasks: sigma0, sigma1, and the
> root server. The current kickstart.cc never loads any sigma1, and the
> spec doesn't say much more about it. So what is sigma1 supposed to do?

I don't know.

> > About that.  Did you check out Neals l4hurd code?
> 
> No, I guess I should. Where exactly can I find that? I browsed the
> l4hurd project at savannah, but didn't find anything there that's
> obviously written by Neal.

On neals web site.
 
> > > For the hurd, the task server needs another grub module for the first
> > > task(s), which could be something like serverboot (processing
> > > additional GRUB modules),
> > 
> > Serverboot is dead, man.
> 
> ;-) The serverboot configuration etc is dead. But some task has to get
> the grub modules needed for booting, and get some tasks running and
> interconnected. Who's going to do that? Not the L4 kernel, not the
> root server, probably not the task server either.

The rootserver it is.

> > > The root server could also support have some ipc call for starting a
> > > new subsystem, later than boot time.
> > 
> > Not at all.
> 
> I don't remeber exactly how I thought when I wrote that. But perhaps
> the right way to get a new subsystem, and start a thread with
> privileges with respect to the root server, is to cooperate with the
> privileged task in your current subsystem. Some coordination on
> subsystem id allocation would be nice, but ok, those are very
> secondary issues for the moment.

Right on, and I think that the easiest thing to say for now is that the
rootserver allows any "privileged" thread to manipulate any other thread in
the system, ie, it does not further check the content of the privileged
system calls.  If there is cross-subsystem interaction required, this is up
to the subsystems themself.

Thanks,
Marcus


-- 
`Rhubarb is no Egyptian god.' GNU      http://www.gnu.org    address@hidden
Marcus Brinkmann              The Hurd http://www.gnu.org/software/hurd/
address@hidden
http://www.marcus-brinkmann.de/