On Mon, Dec 12, 2016 at 10:55:26AM +0800, Paul Goyette wrote: > Hopefully someone(TM) will be able to fix this correctly in the near > future.
I think the include "hack" is fine, but with a real fix I mean: do not map lapic_va at all on machines that do not have one and fix all the callers that eroneously use lapic code. The current code gives an easy way to detect this, while previously this (probably ancient) bug passed unnoticed. Martin