[Top][All Lists]

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

A lot of questions about Mach

From: Marcus Brinkmann
Subject: A lot of questions about Mach
Date: Fri, 29 Jun 2001 15:20:01 -0500
User-agent: Mutt/1.2.5i

On Fri, Jun 29, 2001 at 09:38:03PM +0200, Franck Martin wrote:
> Hi Marcus,
> We’ve no experience in real-time software (just the student level).

The Hurd is not a real time system.  It is neither a goal of the Hurd
project so far, nor is anybody in the Hurd project particularly
experienced in real-time system programming.  (We know the basics, but that's

Real time requirements need to be kernel enforced.  There is RTMach
which is a microkernel with real time support, and the Hurd runs on it.
It will not make use of the real time components, though.

> So,as we need to start, we choose a free software :hurd.
> It will be nice if you’ll be agree to answer our questions.

> Here are our interrogations :
> -     Which platforms are allowed with Hurd ?

Few code in the Hurd is platform dependant.  Again, it is the microkernel
which is.  GNU Mach only runs on i386.  We want ports to other architectures,
butnobody works on them currently.

> -     Could you give us some cross referencer name for development ?do any
> debugging tools existing ?

I don't understand the first question.  There are the Mach docs, and the Hurd
info document.  Canonical source for information about the Hurd is the
Hurd source code.

There is gdb, and you can use it over the serial line with oskit-mach
to debug the kernel itself.

> -     Which kind of communication between task are possible ? (mail box ?,
> shared memory ?, sockets ?)

Mach IPC is port based, the Hurd provides a POSIX interface in glibc.

Shared memory is being worked on.  pflocal and pfinet servers implement
the socket protocols.

> and how do they work ?

Please see the source.

> -     How do hurd manage the task memory allocation ?(pagination ?, dynamic
> allocation ?)

Again, this is a Mach question really.  Mach provides an external pager
interface, and the Hurd implements a default pager.

> -     How many tasks can run in the same time ?

Again a Mach question.  The Hurd does not impose any limit, but the
microkernel might.  You can find the info in Mach documentation or source

> -     What is the time switch between tasks ?

Whatever it is for Mach.

> -     How the task overflow is manage by hurd ?

Again, Mach.  I think it task_create just returns an error value.

> -     How context memory is manage ?

I don't even know what that is.

> -     About the scheduler : Could you explain us how the scheduler is working
> ?(time quantum,
> -     dynamic priority).

I don't know, whatever is in Mach.

> -     In comparaison with VXWorks and PSOS+, what does hurd offer more ?

I don't know either of these other two systems.  The Hurd is a POSIX
compatible operating system licensed under the GPL.  Unlike other
systems, it attempts to make system code as small as possible, and lets
the user extend the system by running their own servers.

> -     What are the mains problems with hurd ?(deadlocks).
> -     Today, which industrials sectors use Hurd (for which applications) ?
> -     Could you give us a basic example of a Hurd program ?
> -     What is your feeling about Hurd ?

I think you should first take a closer look at what the Hurd actually is
before asking more questions.

> We’ve already surf on the web and read GNU sites, but its contains very
> generals informations.

You should have noticed that the Hurd is a multiserver system based on Mach.
Your questions are mostly centered around the basic kernel features Mach
provides, and not particularly targeted at the Hurd.  I invite you to have
a closer look.  A test installation and a few minutes in the source trees
should help.
> We are embarassed :  We don’t speak german ?
> Thanks a lot for your help, cordially,
> Franck, Dominique, Ludovic, Christophe. (french guys)

Next week there is the LSM in Bordeaux (  I will give 
an introductory talk on the Hurd system wed evening or thursday morning
(it's not definitive yet).  You are invited to come and listen, and afterwards
I can answer any more questions you might have.

If you can't come, I recommend reading "New strategy towards OS design" by
Bushnell.  It's on the GNU site (try if you can't find it).


reply via email to

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