Re: x86 minimum processor

To: Mike Hibler <mike@fast.cs.utah.edu>
Subject: Re: x86 minimum processor
From: Matthew Palmer <mjp16@ieee.uow.edu.au>
Date: Wed, 20 Dec 2000 12:05:33 +1100 (EST)
cc: oskit-users@cs.utah.edu
In-Reply-To: <200012200010.RAA23876@fast.cs.utah.edu>
Sender: owner-oskit-users@fast.cs.utah.edu

On Tue, 19 Dec 2000, Mike Hibler wrote:

> > So what is the current minimum processor for x86 OSKit?  I can source a P-II
> > or similar, but if it's a bug in the OSKit (or my fault) I'd like to fix it
> > so I don't have to harass my supervisor for a better test box.
> 
> Well, in theory it should still be a 486.  But obviously we haven't tested
> that in a long time :-)

Aha... <g>

> The most likely culprit is use of the Pentium time stamp counter (via get_tsc
> in proc_reg.h).  I don't think we use it in any default configurations of the
> core libraries, but we have been known to use it in the example kernels.

The really simple kernels work fine - multiboot, hello, etc etc.  The tty
kernel and pingreply aren't nearly so happy.

> Do you get a register dump?  If so, you should be able to use gdb to
> examine what is at the indicated EIP:
> 
> 	x/i 0xNNNNNNNN
> 
> where 0xNNNNNNNN is the address in question.

0x113a9b <oskit_base_irq_handler+11>:	rdtsc

Excellent guess on your part.  Time to update the documentation to recommend
not using a 486 (above and beyond the usual "they're slow, don't use them"
admonishment, of course).

Well, I now have a nice shiny P-II to play with, so this whole issue is now
academic.  Pity.  It's one less thing I can do with a 486.

-- 
-----------------------------------------------------------------------
#include <disclaimer.h>
Matthew Palmer
mjp16@uow.edu.au

References:

Re: x86 minimum processor

From: Mike Hibler <mike@fast.cs.utah.edu>