[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Mach emulation

From: Farid Hajji
Subject: Re: Mach emulation
Date: Wed, 15 Nov 2000 03:37:27 +0100

>   If you write libmom just as a wrapper for L4 API, it wouldnt' be
> fast on other microkernels but L4. Rather, what you aim at is a
> library analogous with libmachuser in glibc.
Is this really true? The problem is how to have the Hurd use only
a minimal set of kernel services. Once this is done, using more
"powerful" (or better: "featureful") ukernels is easy. The reason
for this is simple: Unless the ukernels get even smaller than L4,
they are expected to provide at least the L4 services either directly
or bloated up with more features. In this case, libmom-{newkernel}
could be implemented in terms of direct calls to the more simple
features of the new ukernel that map 1:1 to L4, or adding slightly
more adaptation code.

The only point where I can agree is when a new ukernel provides some
services in a definitive different and much more efficient way
than L4, so that a libmom port would need to go long ways to
"connect" L4-ish services with the new ukernel services. But then,
how likely is this going to be? Nobody knows.

Just one more point: The L4 subset maps without problems to Mach.
A libmom-mach would simply replace L4 syscalls with Mach syscalls,
(using only slighly more bloated messages in some cases, probably
discarding port rights etc...).

To put it another way: L4 or newer ukernels can be considered a
true subset of Mach and L4 is hardly going to become a superset
of a yet smaller/different ukernel (though I can err on that last

The only difficulty I'm seeing with libmom-like-l4-interface is
how to take into account multiprocessor versions of ukernels.
Would it be safe to assume that a multiprocessor ukernel multiplexes
kernel-level threads to the available processors transparently?
This is true for mach, but it may be completely different on
yet-to-be-defined/implemented new multiprocessor-ukernels.


Farid Hajji -- Unix Systems and Network Admin | Phone: +49-2131-67-555
Broicherdorfstr. 83, D-41564 Kaarst, Germany  | address@hidden
- - - - - - - - - - - - - - - - - - - - - - - + - - - - - - - - - - - -
Murphy's Law fails only when you try to demonstrate it, and thus succeeds.

reply via email to

[Prev in Thread] Current Thread [Next in Thread]