Subject: Re: kern/25261: Geode GX1 (id 0x540) clock not broken
To: MINOURA Makoto <email@example.com>
From: Andreas Wrede <firstname.lastname@example.org>
Date: 04/21/2004 02:05:23
MINOURA Makoto wrote:
> |> In <20040420165833.C1D7D1C7A9@woffi.planix.com>
> |> email@example.com wrote:
>>I don't know if there are other versions of the GX1 CPU which needs the fix,
>>my CPU is marked SC1100UFH-266, (C)(M)NSC2002 B1, VS318AB
> According to the Linux sources, this erratic behaviour is
> not due to the processor, but to the old generation I/O
> companion 5510/20. I guess your companion is 5530.
> Since the PCI subsystem is not initialized yet at the time
> the processor is setup, we cannot determine the I/O
> companion. I do not know whether we can defer the errata
> detection after the PCI initialization.
The SC1100 does not have a companion chip - all the (non-video)
functions of the 55x0 are on the SC1100 die.
So the current test in identcpu() should really use the PCI vendor and
device ids of the Cyrix 5510/5520 rather then the fact that we are
running on a GX type CPU?
Hmmm.. Can we create a kernel option to override this behaviour?