multiprocessor support?

[David Chambers]

On 01 Nov 2004 18:07:02 +0000
Colin Paul Adams <colin at colina.demon.co.uk> wrote:

> >>>>> "Gustavo" == Gustavo Seabra <seabra at ksu.edu> writes:
> 
>     Gustavo> Thanks!
>     Gustavo> On Mon, 2004-11-01 at 11:38, Colin Paul Adams wrote:
> 
>     >> >>>>> "Gustavo" == Gustavo Seabra <seabra at ksu.edu> writes:
>     >> 
>     Gustavo> Hi, I am about to install FC2 in one computer that has 2
>     Gustavo> processors.  Currently, it's running on Linux
>     Gustavo> 2.4.18-14smp. Before I make a big mistake, I wonder if
>     Gustavo> anyone here has experience with multiprocessor
>     Gustavo> machines... In other words, does Fedora Core 2 support
>     Gustavo> dual processor?
>     >>  Of course it does.  Personally, I run on a twin Xeon (4
>     >> logical processors) machine.  -- Colin Paul Adams Preston
>     >> Lancashire
> 
> 
> I should have added (but I forgot), that there may be a problem with
> the keyboard - if so, boot in uniprocessor mode, and take one of the
> actions mentioned in bugzilla #????? (I don't remember) to disable
> legacy USB support. In my case, that meant re-compiling the kernel
> without such support, as my BIOS (Phoenix) does not allow disabling
> legacy USB.
> -- 
> Colin Paul Adams
> Preston Lancashire
> 


I have been wrestling with nonfunctional mouse and keyboard on a dual proc
Xeon for a bit now.  You are correct, UP kernels work fine and disabling
legacy USB fixes the problem for SMP - but of course I also have a board
where I can't disable legacy USB in the BIOS.  I found two solutions that
appear to work (I have not found the kernel option to disable legacy USB
yet :-):  (1) Rebuild the kernel RPM with HIGHMEM_64G turned off.  This
means that memory support is limited to 4G though.  (2) Patch i8042.c with
Alan Cox's "i8042_spank_usb()" function (available in a non-used patch on
the current dev kernels). I had to modify the function and where it was
called to get it to work (details if you want them), but it did work. 
To me, both these solutions seem to be skirting around the problem rather
than attacking it at the base, however!

David Chambers
San Diego, CA