Re: Task destruction

[Niels Möller]

Marcus Brinkmann <address@hidden> writes:

> I am not sure about this.  It could give all authorised threads control over
> all tasks (which is what we have in Mach right now, too).

That might work. Although it would be nice if some privileged hurd
server had control over all hurd tasks, while an l4-linux single
server running simultaneously would have control over all linux tasks.
One system should not have control over the other system's tasks
(assuming these operating systems are operating as neighbours,
otherwise one of them would use a proxy task server operated by the
other).

> But I am not sure about how that works in L4. Can tasks die just by
> themselve? I thought they would get an exception, and the exception
> thread would request task death from the (proxy) task server.

Hmm, I suspect you're right. Then the suicide-server, whatever that
turns out to be, is the natural place to put generation of death
notifications.

> All Hurd processes would have to talk to the Hurd proxy task server.

Ok. Then my model would apply to the hurd task server, not to the L4
one. And for a start, it may be simplest to make them one and the
same.

Best regards,
/Niels