In message <20050412211930.GA4081@olib.org>, Richard Rauch writes:
>I would rather suggest *disabling* IOAPIC.
>
>I saw exactly your problem on my AMD64 box a year ago or so.  Somewhere,
>I got the hint of disabling IOAPIC (I never turned on MPACPI).  The problem
>went away.
>
>That doesn't prove that it's the same cause but the symptoms were the same.
>The mouse would attach and work for a while, then it would die at some
>random later time while the system was in use.  (I believe that the whole
>USB port went south.)  This was on a motherboard using an nVidia nForce3
>chipset; others (at least with older nForce3 chipsets) reported similar
>problems.)
>
>(Actually, depending how I had my PCI cards juggled, I recall that sometimes
>the mouse would be okay, but some other devices would fail to configure.  The
>last time that I tried a GENERIC kernel (i.e., with IOAPIC), I still had the
>problem.  I've tried MPACPI, but didn't keep it; I can't remember exactly
>what didn't work with it.  My faint memory is that the problem was the same
>as with IOAPIC.)
>
>
>It might even be worth while removing, or relocating, any PCI cards you
>have.
>

Hmm -- ioapic is disabled in GENERIC now.  I could try a GENERIC kernel 
and see what happens.

		--Prof. Steven M. Bellovin, http://www.cs.columbia.edu/~smb