new kernel for testing

[Alan Cox]

On Wed, Jun 09, 2004 at 09:47:31PM +0200, Arjan van de Ven wrote:
> > Looks to me that you're compiling the usb into the kernel rather than as 
> > modules.  If this is true, this would fix the ps/2 not working on dual 
> > xeon supermicro systems.  Please do seriously consider this 'fix'.
> 
> I do have a plan for trying building the usb host controllers into the
> kernel. But it's post this update to give it a reasonable testing
> time....

I have two machines here that won't boot if you do that. There is also
a much better way to achieve the relevant effect if you know the problems
are USB legacy caused.

Vojtech I think posted a proposed patch that uses pci quirks as an optional
way to simply turn off USB legacy support. That fixes the problem but means
systems that need USB legacy fail - thankfully 2.6.x and the later USB is
now probing very like windows so all the buggy crap that assumes MS windows
works.

The second better way to do it would be to turn off USB legacy, do the initial
probe, turn USB legacy back to its old state and then speak slowly to the
crappy SMM emulation just to check for USB emulation.