linux-mips-fnet
[Top] [All Lists]

Re: Problems booting a 4000SC.

To: linux-mips@fnet.fr
Subject: Re: Problems booting a 4000SC.
From: Ralf Baechle <ralf@cobaltmicro.com>
Date: Thu, 23 Oct 1997 10:58:40 -0700
In-reply-to: <344F713E.4E4C@lsumc.edu>; from Dempsey, Charles on Thu, Oct 23, 1997 at 09:46:06AM -0600
References: <344E3385.23A6@domain.com> <19971022175420.14413@tbird.cobaltmicro.com> <344F713E.4E4C@lsumc.edu>
On Thu, Oct 23, 1997 at 09:46:06AM -0600, Dempsey, Charles wrote:
> The odd things are;
> 
> 1. The SC "processor type" is the same as the PC.

Milo doesn't know that there are subtypes of the R4000 / R4400 processors
and calls them all a R4000PC.

> 2. The cache numbers are the same except for the primary Icache line size, and

Sigh.  I didn't know before that the Magnum PC version caches also have
assymetric linesizes.  Linux doesn't handle that.  Actually I thought
that nobody's using something like that.  I or Thomas will fix that.

If you were wondering why it recognices the machine as "Magnum 4000 SC"
though - the string it's printing is returned by the firmware.

> 3. The secondary cache size is "unknown" in the SC. 

Milo is too dumb to figure the linesize.  In fact it is too dumb to handle
the Magnum's caches.  Luckily it's easy to fix by just transplanting
a couple of the kernel's routines.

> Could this be a boot loader problem? I believe from comparing the floppy
> disk access times on the two machines that the kernel loads but the
> launch fails before video output begins. How does one determine the
> failed process in such a case?

It's hard to trace that, no debugger can do it.  Also at the point of the
boot, where the kernel crashes you don't even have a printf like function
available as Milo killed the ARC firmware and the kernel's console isn't
running yet.  Don't worry ...

  Ralf

<Prev in Thread] Current Thread [Next in Thread>