On Mon, 17 Jun 2002, Manuel Bouyer wrote:

> On Mon, Jun 17, 2002 at 10:47:19AM +0200, Manuel Bouyer wrote:
> > On Mon, Jun 17, 2002 at 01:37:17PM +0900, Makoto Fujiwara wrote:
> > > It looks to me, removing consinit() line from machdep.c
> > > will boot 1.6_BETA2 on StarMax 3000/240 with kbd/screen
> >
> > This doesn't help for me either.
>
> Sorry, I was wrong: this works for me *without* bill's cpu_probe_cache()
> patch. Adding the cpu_probe_cache() make it fail again, in the same
> way.
> Strange.

I really don't understand this. Can you verify that reverting rev 1.4 ->
1.5 on lib/libkern/arch/powerpc/syncicache.c fixes things?

Calling cpu_probe_cache() turns the new code on; it now does nothing if
the probe routine isn't called.

I really don't see why that change would cause this problem. The rev 1.5
code looks exactly like the rev 1.4 code but abstracted to 1) support
different cache line sizes, and 2) do nothing if there is no cache info.

Do you have DDB enabled for this kernel? There are two routines in
consinit(), cninit() and the ddb setup. Let's try and figure out which of
the two causes the problem.

Also, if you can, what happens if you move the consinit() call to later in
that function?

Take care,

Bill